Datasets:

jamescalam
/

llama-2-arxiv-papers-chunked

Dataset card Data Studio Files Files and versions Community

Split (1)

train · 4.84k rows

doi stringlengths 9 10	chunk-id stringlengths 1 4	chunk stringlengths 1 1.57k	id stringlengths 9 10	title stringlengths 10 127	summary stringlengths 581 1.92k	source stringlengths 30 31	authors sequence	categories sequence	comment stringclasses 28 values	journal_ref stringclasses 1 value	primary_category stringclasses 6 values	published stringlengths 8 8	updated stringlengths 8 8	references list
1806.06157	239	In [29], g0is a max pooling operation. The authors show that it allows universal approximation of continuous sets of functions given that the hidden representation (the output of the mapping h()) is of suciently high dimension. We argue that the approximation in (2) can be extended as follows: g(x1; x2; : : : ; xN)f X c2Chc([i2cxi)! (3) whereCis the set of cliques of a graph dened over the item set S,[is the concatenation operator and we chose the sum operator as symmetric function. The input dimension of the non-linearity hc() depends on the size of clique c but maps to a xed output dimension H. In the case where Cis composed of unary cliques only, form (3) decomposes like (2) with the exception of a dierent symmetry operator (sum instead of max pooling). Choosing dierent graphical structures through Swill lead to dierent terms in the summation and allows modeling dierent types of interactions between items in the item set. Note that the interactions between items in (3) are not exclusively modeled	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	240	modeling dierent types of interactions between items in the item set. Note that the interactions between items in (3) are not exclusively modeled through hc(). Indeed, it is interesting to note, that the graphical decomposition provided byCleads to interactions which are dierent from the interactions the same decomposition would provide when used in a probabilistic graphical model, like for instance a Markov Random Field (MRF). In particular, a decomposition into unary terms only, as given in equation (2), does notlead to independence between items, whereas an MRF with unary terms only is equivalent to a distribution over independent random variables. This is a consequence of the global mapping f(), which is dened on the sum over all direct interactions. Higher-order interactions between several items not directly modeled through a non-linearity hc() can eventually be learned by the model through the joint output space of all hc(), provided that the dimensionality Hof this space is high enough to incorporate all interactions. However, whereas the mappinghc() provides a direct model of interactions between pairs of items, learning interactions between two items ( j; k), which are not directly captured through a 8 Baradel et al. visual reasoningmodule	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	241	<latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	242	sha1_base64="fBWZg+aPOFAImknIghA1Bu9U+88=">AAACHXicbVDLSgMxFM3UVx1foy7dBIvgqswUxLoruHFZwdpCp5RM5rYNzSRDkimUoV/ixl9x40LFhRvxb0wfgrYeCBzOOTfJPVHKmTa+/+UU1tY3NreK2+7O7t7+gXd4dK9lpig0qORStSKigTMBDcMMh1aqgCQRh2Y0vJ76zREozaS4M+MUOgnpC9ZjlBgrdb2LMII+EzkFYUBN3BHTGeHYXqGlYKIfhjiRccbBDUHEP7GuV/LL/gx4lQQLUkIL1	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	243	qGlYKIfhjiRccbBDUHEP7GuV/LL/gx4lQQLUkIL1LveRxhLmiV2nHKidTvwU9PJiTKMcpi4YaYhJXRI+tC2VJAEdCefrTfBZ1aJcU8qe4TBM/X3RE4SrcdJZJMJMQO97E3F/7x2ZnrVTs5EmhkQdP5QL+PYSDztCsdMATV8bAmhitm/YjogilDbgXZtCcHyyqukUSlflYPbSqlWXbRRRCfoFJ2jAF2iGrpBddRAFD2gJ/SCXp1H59l5c97n0YKzmDlGf+B8fgM5O6NX</latexit><latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	244	sha1_base64="fBWZg+aPOFAImknIghA1Bu9U+88=">AAACHXicbVDLSgMxFM3UVx1foy7dBIvgqswUxLoruHFZwdpCp5RM5rYNzSRDkimUoV/ixl9x40LFhRvxb0wfgrYeCBzOOTfJPVHKmTa+/+UU1tY3NreK2+7O7t7+gXd4dK9lpig0qORStSKigTMBDcMMh1aqgCQRh2Y0vJ76zREozaS4M+MUOgnpC9ZjlBgrdb2LMII+EzkFYUBN3BHTGeHYXqGlYKIfhjiRccbBDUHEP7GuV/LL/gx4lQQLUkIL1	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	245	qGlYKIfhjiRccbBDUHEP7GuV/LL/gx4lQQLUkIL1LveRxhLmiV2nHKidTvwU9PJiTKMcpi4YaYhJXRI+tC2VJAEdCefrTfBZ1aJcU8qe4TBM/X3RE4SrcdJZJMJMQO97E3F/7x2ZnrVTs5EmhkQdP5QL+PYSDztCsdMATV8bAmhitm/YjogilDbgXZtCcHyyqukUSlflYPbSqlWXbRRRCfoFJ2jAF2iGrpBddRAFD2gJ/SCXp1H59l5c97n0YKzmDlGf+B8fgM5O6NX</latexit><latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	246	sha1_base64="fBWZg+aPOFAImknIghA1Bu9U+88=">AAACHXicbVDLSgMxFM3UVx1foy7dBIvgqswUxLoruHFZwdpCp5RM5rYNzSRDkimUoV/ixl9x40LFhRvxb0wfgrYeCBzOOTfJPVHKmTa+/+UU1tY3NreK2+7O7t7+gXd4dK9lpig0qORStSKigTMBDcMMh1aqgCQRh2Y0vJ76zREozaS4M+MUOgnpC9ZjlBgrdb2LMII+EzkFYUBN3BHTGeHYXqGlYKIfhjiRccbBDUHEP7GuV/LL/gx4lQQLUkIL1	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	247	qGlYKIfhjiRccbBDUHEP7GuV/LL/gx4lQQLUkIL1LveRxhLmiV2nHKidTvwU9PJiTKMcpi4YaYhJXRI+tC2VJAEdCefrTfBZ1aJcU8qe4TBM/X3RE4SrcdJZJMJMQO97E3F/7x2ZnrVTs5EmhkQdP5QL+PYSDztCsdMATV8bAmhitm/YjogilDbgXZtCcHyyqukUSlflYPbSqlWXbRRRCfoFJ2jAF2iGrpBddRAFD2gJ/SCXp1H59l5c97n0YKzmDlGf+B8fgM5O6NX</latexit><latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	248	sha1_base64="fBWZg+aPOFAImknIghA1Bu9U+88=">AAACHXicbVDLSgMxFM3UVx1foy7dBIvgqswUxLoruHFZwdpCp5RM5rYNzSRDkimUoV/ixl9x40LFhRvxb0wfgrYeCBzOOTfJPVHKmTa+/+UU1tY3NreK2+7O7t7+gXd4dK9lpig0qORStSKigTMBDcMMh1aqgCQRh2Y0vJ76zREozaS4M+MUOgnpC9ZjlBgrdb2LMII+EzkFYUBN3BHTGeHYXqGlYKIfhjiRccbBDUHEP7GuV/LL/gx4lQQLUkIL1	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	249	qGlYKIfhjiRccbBDUHEP7GuV/LL/gx4lQQLUkIL1LveRxhLmiV2nHKidTvwU9PJiTKMcpi4YaYhJXRI+tC2VJAEdCefrTfBZ1aJcU8qe4TBM/X3RE4SrcdJZJMJMQO97E3F/7x2ZnrVTs5EmhkQdP5QL+PYSDztCsdMATV8bAmhitm/YjogilDbgXZtCcHyyqukUSlflYPbSqlWXbRRRCfoFJ2jAF2iGrpBddRAFD2gJ/SCXp1H59l5c97n0YKzmDlGf+B8fgM5O6NX</latexit>visual	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	250	reasoningmodule	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	251	<latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	252	sha1_base64="fBWZg+aPOFAImknIghA1Bu9U+88=">AAACHXicbVDLSgMxFM3UVx1foy7dBIvgqswUxLoruHFZwdpCp5RM5rYNzSRDkimUoV/ixl9x40LFhRvxb0wfgrYeCBzOOTfJPVHKmTa+/+UU1tY3NreK2+7O7t7+gXd4dK9lpig0qORStSKigTMBDcMMh1aqgCQRh2Y0vJ76zREozaS4M+MUOgnpC9ZjlBgrdb2LMII+EzkFYUBN3BHTGeHYXqGlYKIfhjiRccbBDUHEP7GuV/LL/gx4lQQLUkIL1	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	253	qGlYKIfhjiRccbBDUHEP7GuV/LL/gx4lQQLUkIL1LveRxhLmiV2nHKidTvwU9PJiTKMcpi4YaYhJXRI+tC2VJAEdCefrTfBZ1aJcU8qe4TBM/X3RE4SrcdJZJMJMQO97E3F/7x2ZnrVTs5EmhkQdP5QL+PYSDztCsdMATV8bAmhitm/YjogilDbgXZtCcHyyqukUSlflYPbSqlWXbRRRCfoFJ2jAF2iGrpBddRAFD2gJ/SCXp1H59l5c97n0YKzmDlGf+B8fgM5O6NX</latexit><latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	254	sha1_base64="fBWZg+aPOFAImknIghA1Bu9U+88=">AAACHXicbVDLSgMxFM3UVx1foy7dBIvgqswUxLoruHFZwdpCp5RM5rYNzSRDkimUoV/ixl9x40LFhRvxb0wfgrYeCBzOOTfJPVHKmTa+/+UU1tY3NreK2+7O7t7+gXd4dK9lpig0qORStSKigTMBDcMMh1aqgCQRh2Y0vJ76zREozaS4M+MUOgnpC9ZjlBgrdb2LMII+EzkFYUBN3BHTGeHYXqGlYKIfhjiRccbBDUHEP7GuV/LL/gx4lQQLUkIL1	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	255	qGlYKIfhjiRccbBDUHEP7GuV/LL/gx4lQQLUkIL1LveRxhLmiV2nHKidTvwU9PJiTKMcpi4YaYhJXRI+tC2VJAEdCefrTfBZ1aJcU8qe4TBM/X3RE4SrcdJZJMJMQO97E3F/7x2ZnrVTs5EmhkQdP5QL+PYSDztCsdMATV8bAmhitm/YjogilDbgXZtCcHyyqukUSlflYPbSqlWXbRRRCfoFJ2jAF2iGrpBddRAFD2gJ/SCXp1H59l5c97n0YKzmDlGf+B8fgM5O6NX</latexit><latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	256	sha1_base64="fBWZg+aPOFAImknIghA1Bu9U+88=">AAACHXicbVDLSgMxFM3UVx1foy7dBIvgqswUxLoruHFZwdpCp5RM5rYNzSRDkimUoV/ixl9x40LFhRvxb0wfgrYeCBzOOTfJPVHKmTa+/+UU1tY3NreK2+7O7t7+gXd4dK9lpig0qORStSKigTMBDcMMh1aqgCQRh2Y0vJ76zREozaS4M+MUOgnpC9ZjlBgrdb2LMII+EzkFYUBN3BHTGeHYXqGlYKIfhjiRccbBDUHEP7GuV/LL/gx4lQQLUkIL1	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	257	qGlYKIfhjiRccbBDUHEP7GuV/LL/gx4lQQLUkIL1LveRxhLmiV2nHKidTvwU9PJiTKMcpi4YaYhJXRI+tC2VJAEdCefrTfBZ1aJcU8qe4TBM/X3RE4SrcdJZJMJMQO97E3F/7x2ZnrVTs5EmhkQdP5QL+PYSDztCsdMATV8bAmhitm/YjogilDbgXZtCcHyyqukUSlflYPbSqlWXbRRRCfoFJ2jAF2iGrpBddRAFD2gJ/SCXp1H59l5c97n0YKzmDlGf+B8fgM5O6NX</latexit><latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	258	sha1_base64="fBWZg+aPOFAImknIghA1Bu9U+88=">AAACHXicbVDLSgMxFM3UVx1foy7dBIvgqswUxLoruHFZwdpCp5RM5rYNzSRDkimUoV/ixl9x40LFhRvxb0wfgrYeCBzOOTfJPVHKmTa+/+UU1tY3NreK2+7O7t7+gXd4dK9lpig0qORStSKigTMBDcMMh1aqgCQRh2Y0vJ76zREozaS4M+MUOgnpC9ZjlBgrdb2LMII+EzkFYUBN3BHTGeHYXqGlYKIfhjiRccbBDUHEP7GuV/LL/gx4lQQLUkIL1	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	259	qGlYKIfhjiRccbBDUHEP7GuV/LL/gx4lQQLUkIL1LveRxhLmiV2nHKidTvwU9PJiTKMcpi4YaYhJXRI+tC2VJAEdCefrTfBZ1aJcU8qe4TBM/X3RE4SrcdJZJMJMQO97E3F/7x2ZnrVTs5EmhkQdP5QL+PYSDztCsdMATV8bAmhitm/YjogilDbgXZtCcHyyqukUSlflYPbSqlWXbRRRCfoFJ2jAF2iGrpBddRAFD2gJ/SCXp1H59l5c97n0YKzmDlGf+B8fgM5O6NX</latexit>random	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	260	framet02[0...t2]	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	261	<latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	262	sha1_base64="3hmJJGXFri/4Fn9gRqnNsEIPzF4=">AAACJ3icbVDLSsQwFE19W1+jLt0ER9GNQzsbdaXgxqWCo8K0DGl6OwbTpCS3wlDmc9z4K25EVHTpn5gZR/B1IHA45x5y70kKKSwGwZs3Nj4xOTU9M+vPzS8sLtWWV86tLg2HFtdSm8uEWZBCQQsFSrgsDLA8kXCRXB8N/IsbMFZodYa9AuKcdZXIBGfopE7tIEqgK1TFQSGYvm+YSnVOM8NyoBu4FQnVDiKZarQUq51+M97wI1DpV6BTqweNYAj6l	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	263	Bu4FQnVDiKZarQUq51+M97wI1DpV6BTqweNYAj6l4QjUicjnHRqj1GqeZm7OJfM2nYYFBhXzKDgEvp+VFooGL9mXWg7qtweNq6Gh/bpplNSmmnjnkI6VL8nKpZb28sTN5kzvLK/vYH4n9cuMduLK6GKEkHxz4+yUlLUdNAaTYUBjrLnCONGuF0pv2KGcdeB9V0J4e+T/5JWs7HfCE+b9cPtURszZI2sk20Skl1ySI7JCWkRTm7JPXkiz96d9+C9eK+fo2PeKLNKfsB7/wD5faYK</latexit><latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	264	sha1_base64="3hmJJGXFri/4Fn9gRqnNsEIPzF4=">AAACJ3icbVDLSsQwFE19W1+jLt0ER9GNQzsbdaXgxqWCo8K0DGl6OwbTpCS3wlDmc9z4K25EVHTpn5gZR/B1IHA45x5y70kKKSwGwZs3Nj4xOTU9M+vPzS8sLtWWV86tLg2HFtdSm8uEWZBCQQsFSrgsDLA8kXCRXB8N/IsbMFZodYa9AuKcdZXIBGfopE7tIEqgK1TFQSGYvm+YSnVOM8NyoBu4FQnVDiKZarQUq51+M97wI1DpV6BTqweNYAj6l	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	265	Bu4FQnVDiKZarQUq51+M97wI1DpV6BTqweNYAj6l4QjUicjnHRqj1GqeZm7OJfM2nYYFBhXzKDgEvp+VFooGL9mXWg7qtweNq6Gh/bpplNSmmnjnkI6VL8nKpZb28sTN5kzvLK/vYH4n9cuMduLK6GKEkHxz4+yUlLUdNAaTYUBjrLnCONGuF0pv2KGcdeB9V0J4e+T/5JWs7HfCE+b9cPtURszZI2sk20Skl1ySI7JCWkRTm7JPXkiz96d9+C9eK+fo2PeKLNKfsB7/wD5faYK</latexit><latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	266	sha1_base64="3hmJJGXFri/4Fn9gRqnNsEIPzF4=">AAACJ3icbVDLSsQwFE19W1+jLt0ER9GNQzsbdaXgxqWCo8K0DGl6OwbTpCS3wlDmc9z4K25EVHTpn5gZR/B1IHA45x5y70kKKSwGwZs3Nj4xOTU9M+vPzS8sLtWWV86tLg2HFtdSm8uEWZBCQQsFSrgsDLA8kXCRXB8N/IsbMFZodYa9AuKcdZXIBGfopE7tIEqgK1TFQSGYvm+YSnVOM8NyoBu4FQnVDiKZarQUq51+M97wI1DpV6BTqweNYAj6l	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	267	Bu4FQnVDiKZarQUq51+M97wI1DpV6BTqweNYAj6l4QjUicjnHRqj1GqeZm7OJfM2nYYFBhXzKDgEvp+VFooGL9mXWg7qtweNq6Gh/bpplNSmmnjnkI6VL8nKpZb28sTN5kzvLK/vYH4n9cuMduLK6GKEkHxz4+yUlLUdNAaTYUBjrLnCONGuF0pv2KGcdeB9V0J4e+T/5JWs7HfCE+b9cPtURszZI2sk20Skl1ySI7JCWkRTm7JPXkiz96d9+C9eK+fo2PeKLNKfsB7/wD5faYK</latexit><latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	268	sha1_base64="3hmJJGXFri/4Fn9gRqnNsEIPzF4=">AAACJ3icbVDLSsQwFE19W1+jLt0ER9GNQzsbdaXgxqWCo8K0DGl6OwbTpCS3wlDmc9z4K25EVHTpn5gZR/B1IHA45x5y70kKKSwGwZs3Nj4xOTU9M+vPzS8sLtWWV86tLg2HFtdSm8uEWZBCQQsFSrgsDLA8kXCRXB8N/IsbMFZodYa9AuKcdZXIBGfopE7tIEqgK1TFQSGYvm+YSnVOM8NyoBu4FQnVDiKZarQUq51+M97wI1DpV6BTqweNYAj6l	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	269	Bu4FQnVDiKZarQUq51+M97wI1DpV6BTqweNYAj6l4QjUicjnHRqj1GqeZm7OJfM2nYYFBhXzKDgEvp+VFooGL9mXWg7qtweNq6Gh/bpplNSmmnjnkI6VL8nKpZb28sTN5kzvLK/vYH4n9cuMduLK6GKEkHxz4+yUlLUdNAaTYUBjrLnCONGuF0pv2KGcdeB9V0J4e+T/5JWs7HfCE+b9cPtURszZI2sk20Skl1ySI7JCWkRTm7JPXkiz96d9+C9eK+fo2PeKLNKfsB7/wD5faYK</latexit>random	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	270	framet002[0...t1]	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	271	<latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	272	sha1_base64="A/qj6iEXV6zX9h9ZfaxCC8IJQkw=">AAACKHicbVDLSgMxFM34dnxVXboJVtGNZcaNuiu4cVnBqtAZSiZzpwYzyZDcEcrQ33Hjr7hRUOnWLzGtFXwdCBzOuYfce5JCCotBMPSmpmdm5+YXFv2l5ZXVtdr6xqXVpeHQ5lpqc50wC1IoaKNACdeFAZYnEq6S29ORf3UHxgqtLrBfQJyznhKZ4Ayd1K01owR6QlUcFIIZ+IapVOc0MywHuoN7e5FQnSCSqUZLsToYhPGOH4FKvxLdWj1oBGPQv	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	273	oN7e5FQnSCSqUZLsToYhPGOH4FKvxLdWj1oBGPQvySckDqZoNWtPUep5mXu4lwyazthUGBcMYOCSxj4UWmhYPyW9aDjqHKL2LgaXzqgu05JaaaNewrpWP2eqFhubT9P3GTO8Mb+9kbif16nxOw4roQqSgTFPz/KSklR01FtNBUGOMq+I4wb4Xal/IYZxl0H1nclhL9P/kvah42TRnh+WG/uT9pYIFtkm+yTkByRJjkjLdImnNyTR/JCXr0H78l784afo1PeJLNJfsB7/wBovKY6</latexit><latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	274	sha1_base64="A/qj6iEXV6zX9h9ZfaxCC8IJQkw=">AAACKHicbVDLSgMxFM34dnxVXboJVtGNZcaNuiu4cVnBqtAZSiZzpwYzyZDcEcrQ33Hjr7hRUOnWLzGtFXwdCBzOuYfce5JCCotBMPSmpmdm5+YXFv2l5ZXVtdr6xqXVpeHQ5lpqc50wC1IoaKNACdeFAZYnEq6S29ORf3UHxgqtLrBfQJyznhKZ4Ayd1K01owR6QlUcFIIZ+IapVOc0MywHuoN7e5FQnSCSqUZLsToYhPGOH4FKvxLdWj1oBGPQv	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	275	oN7e5FQnSCSqUZLsToYhPGOH4FKvxLdWj1oBGPQvySckDqZoNWtPUep5mXu4lwyazthUGBcMYOCSxj4UWmhYPyW9aDjqHKL2LgaXzqgu05JaaaNewrpWP2eqFhubT9P3GTO8Mb+9kbif16nxOw4roQqSgTFPz/KSklR01FtNBUGOMq+I4wb4Xal/IYZxl0H1nclhL9P/kvah42TRnh+WG/uT9pYIFtkm+yTkByRJjkjLdImnNyTR/JCXr0H78l784afo1PeJLNJfsB7/wBovKY6</latexit><latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	276	sha1_base64="A/qj6iEXV6zX9h9ZfaxCC8IJQkw=">AAACKHicbVDLSgMxFM34dnxVXboJVtGNZcaNuiu4cVnBqtAZSiZzpwYzyZDcEcrQ33Hjr7hRUOnWLzGtFXwdCBzOuYfce5JCCotBMPSmpmdm5+YXFv2l5ZXVtdr6xqXVpeHQ5lpqc50wC1IoaKNACdeFAZYnEq6S29ORf3UHxgqtLrBfQJyznhKZ4Ayd1K01owR6QlUcFIIZ+IapVOc0MywHuoN7e5FQnSCSqUZLsToYhPGOH4FKvxLdWj1oBGPQv	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	277	oN7e5FQnSCSqUZLsToYhPGOH4FKvxLdWj1oBGPQvySckDqZoNWtPUep5mXu4lwyazthUGBcMYOCSxj4UWmhYPyW9aDjqHKL2LgaXzqgu05JaaaNewrpWP2eqFhubT9P3GTO8Mb+9kbif16nxOw4roQqSgTFPz/KSklR01FtNBUGOMq+I4wb4Xal/IYZxl0H1nclhL9P/kvah42TRnh+WG/uT9pYIFtkm+yTkByRJjkjLdImnNyTR/JCXr0H78l784afo1PeJLNJfsB7/wBovKY6</latexit><latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	278	sha1_base64="A/qj6iEXV6zX9h9ZfaxCC8IJQkw=">AAACKHicbVDLSgMxFM34dnxVXboJVtGNZcaNuiu4cVnBqtAZSiZzpwYzyZDcEcrQ33Hjr7hRUOnWLzGtFXwdCBzOuYfce5JCCotBMPSmpmdm5+YXFv2l5ZXVtdr6xqXVpeHQ5lpqc50wC1IoaKNACdeFAZYnEq6S29ORf3UHxgqtLrBfQJyznhKZ4Ayd1K01owR6QlUcFIIZ+IapVOc0MywHuoN7e5FQnSCSqUZLsToYhPGOH4FKvxLdWj1oBGPQv	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	279	oN7e5FQnSCSqUZLsToYhPGOH4FKvxLdWj1oBGPQvySckDqZoNWtPUep5mXu4lwyazthUGBcMYOCSxj4UWmhYPyW9aDjqHKL2LgaXzqgu05JaaaNewrpWP2eqFhubT9P3GTO8Mb+9kbif16nxOw4roQqSgTFPz/KSklR01FtNBUGOMq+I4wb4Xal/IYZxl0H1nclhL9P/kvah42TRnh+WG/uT9pYIFtkm+yTkByRJjkjLdImnNyTR/JCXr0H78l784afo1PeJLNJfsB7/wBovKY6</latexit>framet1	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	280	<latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	281	sha1_base64="QdD1TILqtKPiRr85PfCrEm51r9o=">AAACEXicbVDLSsNAFJ3UV42vqEs3g61QF5akG3VXcOOygrGFppTJ9KYdOpmEmYlQQr/Bjb/ixoWKW3fu/BunD0FbDwwczrmHO/eEKWdKu+6XVVhZXVvfKG7aW9s7u3vO/sGdSjJJwacJT2QrJAo4E+Brpjm0UgkkDjk0w+HVxG/eg1QsEbd6lEInJn3BIkaJNlLXOQ1C6DORUxAa5NiOJIkBl3V+NvbKdgCi92N1nZJbdafAy8SbkxKao9F1PoNeQ	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	282	3V+NvbKdgCi92N1nZJbdafAy8SbkxKao9F1PoNeQrPYxCknSrU9N9WdnEjNKIexHWQKUkKHpA9tQ4VZrDr59KQxPjFKD0eJNE9oPFV/J3ISKzWKQzMZEz1Qi95E/M9rZzq66ORMpJkGQWeLooxjneBJP7jHJFDNR4YQKpn5K6YDIgk1HSjblOAtnrxM/Fr1surd1Er1yryNIjpCx6iCPHSO6ugaNZCPKHpAT+gFvVqP1rP1Zr3PRgvWPHOI/sD6+AYPMJ00</latexit><latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	283	sha1_base64="QdD1TILqtKPiRr85PfCrEm51r9o=">AAACEXicbVDLSsNAFJ3UV42vqEs3g61QF5akG3VXcOOygrGFppTJ9KYdOpmEmYlQQr/Bjb/ixoWKW3fu/BunD0FbDwwczrmHO/eEKWdKu+6XVVhZXVvfKG7aW9s7u3vO/sGdSjJJwacJT2QrJAo4E+Brpjm0UgkkDjk0w+HVxG/eg1QsEbd6lEInJn3BIkaJNlLXOQ1C6DORUxAa5NiOJIkBl3V+NvbKdgCi92N1nZJbdafAy8SbkxKao9F1PoNeQ	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	284	3V+NvbKdgCi92N1nZJbdafAy8SbkxKao9F1PoNeQrPYxCknSrU9N9WdnEjNKIexHWQKUkKHpA9tQ4VZrDr59KQxPjFKD0eJNE9oPFV/J3ISKzWKQzMZEz1Qi95E/M9rZzq66ORMpJkGQWeLooxjneBJP7jHJFDNR4YQKpn5K6YDIgk1HSjblOAtnrxM/Fr1surd1Er1yryNIjpCx6iCPHSO6ugaNZCPKHpAT+gFvVqP1rP1Zr3PRgvWPHOI/sD6+AYPMJ00</latexit><latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	285	sha1_base64="QdD1TILqtKPiRr85PfCrEm51r9o=">AAACEXicbVDLSsNAFJ3UV42vqEs3g61QF5akG3VXcOOygrGFppTJ9KYdOpmEmYlQQr/Bjb/ixoWKW3fu/BunD0FbDwwczrmHO/eEKWdKu+6XVVhZXVvfKG7aW9s7u3vO/sGdSjJJwacJT2QrJAo4E+Brpjm0UgkkDjk0w+HVxG/eg1QsEbd6lEInJn3BIkaJNlLXOQ1C6DORUxAa5NiOJIkBl3V+NvbKdgCi92N1nZJbdafAy8SbkxKao9F1PoNeQ	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	286	3V+NvbKdgCi92N1nZJbdafAy8SbkxKao9F1PoNeQrPYxCknSrU9N9WdnEjNKIexHWQKUkKHpA9tQ4VZrDr59KQxPjFKD0eJNE9oPFV/J3ISKzWKQzMZEz1Qi95E/M9rZzq66ORMpJkGQWeLooxjneBJP7jHJFDNR4YQKpn5K6YDIgk1HSjblOAtnrxM/Fr1surd1Er1yryNIjpCx6iCPHSO6ugaNZCPKHpAT+gFvVqP1rP1Zr3PRgvWPHOI/sD6+AYPMJ00</latexit><latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	287	sha1_base64="QdD1TILqtKPiRr85PfCrEm51r9o=">AAACEXicbVDLSsNAFJ3UV42vqEs3g61QF5akG3VXcOOygrGFppTJ9KYdOpmEmYlQQr/Bjb/ixoWKW3fu/BunD0FbDwwczrmHO/eEKWdKu+6XVVhZXVvfKG7aW9s7u3vO/sGdSjJJwacJT2QrJAo4E+Brpjm0UgkkDjk0w+HVxG/eg1QsEbd6lEInJn3BIkaJNlLXOQ1C6DORUxAa5NiOJIkBl3V+NvbKdgCi92N1nZJbdafAy8SbkxKao9F1PoNeQ	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	288	3V+NvbKdgCi92N1nZJbdafAy8SbkxKao9F1PoNeQrPYxCknSrU9N9WdnEjNKIexHWQKUkKHpA9tQ4VZrDr59KQxPjFKD0eJNE9oPFV/J3ISKzWKQzMZEz1Qi95E/M9rZzq66ORMpJkGQWeLooxjneBJP7jHJFDNR4YQKpn5K6YDIgk1HSjblOAtnrxM/Fr1surd1Er1yryNIjpCx6iCPHSO6ugaNZCPKHpAT+gFvVqP1rP1Zr3PRgvWPHOI/sD6+AYPMJ00</latexit>framet	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	289	<latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	290	sha1_base64="BD8grwOxxSmQYxdsSCDxptyziy4=">AAACDXicbVBNS8NAFNz4WeNX1aOXxbbQU0l6UW8FLx4rWFtoQtlsXtqlm03Y3Qgl9Bd48a948aDi1bs3/43bNoK2DiwMM294+yZIOVPacb6stfWNza3t0o69u7d/cFg+Or5TSSYpdGjCE9kLiALOBHQ00xx6qQQSBxy6wfhq5nfvQSqWiFs9ScGPyVCwiFGijTQo17wAhkzkFIQGObUjSWLAVV21PRDhjzwoV5yGMwdeJW5BKqhAe1D+9MKEZrGJU	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	291	V21PRDhjzwoV5yGMwdeJW5BKqhAe1D+9MKEZrGJU06U6rtOqv2cSM0oh6ntZQpSQsdkCH1DhVmq/Hx+zhTXjBLiKJHmCY3n6u9ETmKlJnFgJmOiR2rZm4n/ef1MRxd+zkSaaRB0sSjKONYJnnWDQyaBaj4xhFDJzF8xHRFJqOlA2aYEd/nkVdJpNi4b7k2z0qoXbZTQKTpDdeSic9RC16iNOoiiB/SEXtCr9Wg9W2/W+2J0zSoyJ+gPrI9vPTWbtg==</latexit><latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	292	sha1_base64="BD8grwOxxSmQYxdsSCDxptyziy4=">AAACDXicbVBNS8NAFNz4WeNX1aOXxbbQU0l6UW8FLx4rWFtoQtlsXtqlm03Y3Qgl9Bd48a948aDi1bs3/43bNoK2DiwMM294+yZIOVPacb6stfWNza3t0o69u7d/cFg+Or5TSSYpdGjCE9kLiALOBHQ00xx6qQQSBxy6wfhq5nfvQSqWiFs9ScGPyVCwiFGijTQo17wAhkzkFIQGObUjSWLAVV21PRDhjzwoV5yGMwdeJW5BKqhAe1D+9MKEZrGJU	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	293	V21PRDhjzwoV5yGMwdeJW5BKqhAe1D+9MKEZrGJU06U6rtOqv2cSM0oh6ntZQpSQsdkCH1DhVmq/Hx+zhTXjBLiKJHmCY3n6u9ETmKlJnFgJmOiR2rZm4n/ef1MRxd+zkSaaRB0sSjKONYJnnWDQyaBaj4xhFDJzF8xHRFJqOlA2aYEd/nkVdJpNi4b7k2z0qoXbZTQKTpDdeSic9RC16iNOoiiB/SEXtCr9Wg9W2/W+2J0zSoyJ+gPrI9vPTWbtg==</latexit><latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	294	sha1_base64="BD8grwOxxSmQYxdsSCDxptyziy4=">AAACDXicbVBNS8NAFNz4WeNX1aOXxbbQU0l6UW8FLx4rWFtoQtlsXtqlm03Y3Qgl9Bd48a948aDi1bs3/43bNoK2DiwMM294+yZIOVPacb6stfWNza3t0o69u7d/cFg+Or5TSSYpdGjCE9kLiALOBHQ00xx6qQQSBxy6wfhq5nfvQSqWiFs9ScGPyVCwiFGijTQo17wAhkzkFIQGObUjSWLAVV21PRDhjzwoV5yGMwdeJW5BKqhAe1D+9MKEZrGJU	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	295	V21PRDhjzwoV5yGMwdeJW5BKqhAe1D+9MKEZrGJU06U6rtOqv2cSM0oh6ntZQpSQsdkCH1DhVmq/Hx+zhTXjBLiKJHmCY3n6u9ETmKlJnFgJmOiR2rZm4n/ef1MRxd+zkSaaRB0sSjKONYJnnWDQyaBaj4xhFDJzF8xHRFJqOlA2aYEd/nkVdJpNi4b7k2z0qoXbZTQKTpDdeSic9RC16iNOoiiB/SEXtCr9Wg9W2/W+2J0zSoyJ+gPrI9vPTWbtg==</latexit><latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	296	sha1_base64="BD8grwOxxSmQYxdsSCDxptyziy4=">AAACDXicbVBNS8NAFNz4WeNX1aOXxbbQU0l6UW8FLx4rWFtoQtlsXtqlm03Y3Qgl9Bd48a948aDi1bs3/43bNoK2DiwMM294+yZIOVPacb6stfWNza3t0o69u7d/cFg+Or5TSSYpdGjCE9kLiALOBHQ00xx6qQQSBxy6wfhq5nfvQSqWiFs9ScGPyVCwiFGijTQo17wAhkzkFIQGObUjSWLAVV21PRDhjzwoV5yGMwdeJW5BKqhAe1D+9MKEZrGJU	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	297	V21PRDhjzwoV5yGMwdeJW5BKqhAe1D+9MKEZrGJU06U6rtOqv2cSM0oh6ntZQpSQsdkCH1DhVmq/Hx+zhTXjBLiKJHmCY3n6u9ETmKlJnFgJmOiR2rZm4n/ef1MRxd+zkSaaRB0sSjKONYJnnWDQyaBaj4xhFDJzF8xHRFJqOlA2aYEd/nkVdJpNi4b7k2z0qoXbZTQKTpDdeSic9RC16iNOoiiB/SEXtCr9Wg9W2/W+2J0zSoyJ+gPrI9vPTWbtg==</latexit>video	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	298	stream<latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	299	sha1_base64="tv8v5jh4gPs3lF/UcaOyaYFtsvY=">AAACEHicbVDLSgMxFM34rOOr6tJNsAhdlZlu1F3BjcsKji10Sslk7rShmWRIMoUy9Bfc+CtuXKi4denOvzF9CNp6IHA4554k90QZZ9p43peztr6xubVd2nF39/YPDstHx/da5opCQCWXqh0RDZwJCAwzHNqZApJGHFrR8Hrqt0agNJPizowz6KakL1jCKDFW6pWrYQR9JgoKwoCauCMWg8TaTO9wQxDxj9MrV7yaNwNeJf6CVNACzV75M4wlzVMbp	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	300	O9wQxDxj9MrV7yaNwNeJf6CVNACzV75M4wlzVMbp5xo3fG9zHQLogyjHCZumGvICB2SPnQsFSQF3S1mG03wuVVinEhljzB4pv5OFCTVepxGdjIlZqCXvan4n9fJTXLZLZjIcgOCzh9Kco6NxNN6cMwUUMPHlhCqmP0rpgOiCLUdaNeW4C+vvEqCeu2q5t/WK43qoo0SOkVnqIp8dIEa6AY1UYAoekBP6AW9Oo/Os/PmvM9H15xF5gT9gfPxDbN+nbA=</latexit><latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	301	sha1_base64="tv8v5jh4gPs3lF/UcaOyaYFtsvY=">AAACEHicbVDLSgMxFM34rOOr6tJNsAhdlZlu1F3BjcsKji10Sslk7rShmWRIMoUy9Bfc+CtuXKi4denOvzF9CNp6IHA4554k90QZZ9p43peztr6xubVd2nF39/YPDstHx/da5opCQCWXqh0RDZwJCAwzHNqZApJGHFrR8Hrqt0agNJPizowz6KakL1jCKDFW6pWrYQR9JgoKwoCauCMWg8TaTO9wQxDxj9MrV7yaNwNeJf6CVNACzV75M4wlzVMbp	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	302	O9wQxDxj9MrV7yaNwNeJf6CVNACzV75M4wlzVMbp5xo3fG9zHQLogyjHCZumGvICB2SPnQsFSQF3S1mG03wuVVinEhljzB4pv5OFCTVepxGdjIlZqCXvan4n9fJTXLZLZjIcgOCzh9Kco6NxNN6cMwUUMPHlhCqmP0rpgOiCLUdaNeW4C+vvEqCeu2q5t/WK43qoo0SOkVnqIp8dIEa6AY1UYAoekBP6AW9Oo/Os/PmvM9H15xF5gT9gfPxDbN+nbA=</latexit><latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	303	sha1_base64="tv8v5jh4gPs3lF/UcaOyaYFtsvY=">AAACEHicbVDLSgMxFM34rOOr6tJNsAhdlZlu1F3BjcsKji10Sslk7rShmWRIMoUy9Bfc+CtuXKi4denOvzF9CNp6IHA4554k90QZZ9p43peztr6xubVd2nF39/YPDstHx/da5opCQCWXqh0RDZwJCAwzHNqZApJGHFrR8Hrqt0agNJPizowz6KakL1jCKDFW6pWrYQR9JgoKwoCauCMWg8TaTO9wQxDxj9MrV7yaNwNeJf6CVNACzV75M4wlzVMbp	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	304	O9wQxDxj9MrV7yaNwNeJf6CVNACzV75M4wlzVMbp5xo3fG9zHQLogyjHCZumGvICB2SPnQsFSQF3S1mG03wuVVinEhljzB4pv5OFCTVepxGdjIlZqCXvan4n9fJTXLZLZjIcgOCzh9Kco6NxNN6cMwUUMPHlhCqmP0rpgOiCLUdaNeW4C+vvEqCeu2q5t/WK43qoo0SOkVnqIp8dIEa6AY1UYAoekBP6AW9Oo/Os/PmvM9H15xF5gT9gfPxDbN+nbA=</latexit><latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	305	sha1_base64="tv8v5jh4gPs3lF/UcaOyaYFtsvY=">AAACEHicbVDLSgMxFM34rOOr6tJNsAhdlZlu1F3BjcsKji10Sslk7rShmWRIMoUy9Bfc+CtuXKi4denOvzF9CNp6IHA4554k90QZZ9p43peztr6xubVd2nF39/YPDstHx/da5opCQCWXqh0RDZwJCAwzHNqZApJGHFrR8Hrqt0agNJPizowz6KakL1jCKDFW6pWrYQR9JgoKwoCauCMWg8TaTO9wQxDxj9MrV7yaNwNeJf6CVNACzV75M4wlzVMbp	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	306	O9wQxDxj9MrV7yaNwNeJf6CVNACzV75M4wlzVMbp5xo3fG9zHQLogyjHCZumGvICB2SPnQsFSQF3S1mG03wuVVinEhljzB4pv5OFCTVepxGdjIlZqCXvan4n9fJTXLZLZjIcgOCzh9Kco6NxNN6cMwUUMPHlhCqmP0rpgOiCLUdaNeW4C+vvEqCeu2q5t/WK43qoo0SOkVnqIp8dIEa6AY1UYAoekBP6AW9Oo/Os/PmvM9H15xF5gT9gfPxDbN+nbA=</latexit>	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	307	RNN	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	308	<latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	309	sha1_base64="WNqRVHooK3MRbsx3jlRlp53Ez9s=">AAACB3icbVBNS8NAFNzUrxq/oh49GCyCp5L0Yr0VvHgqVYwttKFsNi/t0s0m7G6EEnr04l/x4kHFq3/Bm//GbRtBWwcWhpk3vH0TpIxK5ThfRmlldW19o7xpbm3v7O5Z+wd3MskEAY8kLBGdAEtglIOnqGLQSQXgOGDQDkaXU799D0LShN+qcQp+jAecRpRgpaW+ddwLYEB5ToArEBPzptk0e8DDH6FvVZyqM4O9TNyCVFCBVt/67IUJyWIdJwxL2	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	310	8DDH6FvVZyqM4O9TNyCVFCBVt/67IUJyWIdJwxL2XWdVPk5FooSBhOzl0lIMRnhAXQ15TgG6eezQyb2qVZCO0qEflzZM/V3IsexlOM40JMxVkO56E3F/7xupqK6n1OeZgo4mS+KMmarxJ62YodUAFFsrAkmguq/2mSIBSa6A2nqEtzFk5eJV6teVN3rWqVRL9oooyN0gs6Qi85RA12hFvIQQQ/oCb2gV+PReDbejPf5aMkoMofoD4yPbx1cmZE=</latexit><latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	311	sha1_base64="WNqRVHooK3MRbsx3jlRlp53Ez9s=">AAACB3icbVBNS8NAFNzUrxq/oh49GCyCp5L0Yr0VvHgqVYwttKFsNi/t0s0m7G6EEnr04l/x4kHFq3/Bm//GbRtBWwcWhpk3vH0TpIxK5ThfRmlldW19o7xpbm3v7O5Z+wd3MskEAY8kLBGdAEtglIOnqGLQSQXgOGDQDkaXU799D0LShN+qcQp+jAecRpRgpaW+ddwLYEB5ToArEBPzptk0e8DDH6FvVZyqM4O9TNyCVFCBVt/67IUJyWIdJwxL2	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	312	8DDH6FvVZyqM4O9TNyCVFCBVt/67IUJyWIdJwxL2XWdVPk5FooSBhOzl0lIMRnhAXQ15TgG6eezQyb2qVZCO0qEflzZM/V3IsexlOM40JMxVkO56E3F/7xupqK6n1OeZgo4mS+KMmarxJ62YodUAFFsrAkmguq/2mSIBSa6A2nqEtzFk5eJV6teVN3rWqVRL9oooyN0gs6Qi85RA12hFvIQQQ/oCb2gV+PReDbejPf5aMkoMofoD4yPbx1cmZE=</latexit><latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	313	sha1_base64="WNqRVHooK3MRbsx3jlRlp53Ez9s=">AAACB3icbVBNS8NAFNzUrxq/oh49GCyCp5L0Yr0VvHgqVYwttKFsNi/t0s0m7G6EEnr04l/x4kHFq3/Bm//GbRtBWwcWhpk3vH0TpIxK5ThfRmlldW19o7xpbm3v7O5Z+wd3MskEAY8kLBGdAEtglIOnqGLQSQXgOGDQDkaXU799D0LShN+qcQp+jAecRpRgpaW+ddwLYEB5ToArEBPzptk0e8DDH6FvVZyqM4O9TNyCVFCBVt/67IUJyWIdJwxL2	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	314	8DDH6FvVZyqM4O9TNyCVFCBVt/67IUJyWIdJwxL2XWdVPk5FooSBhOzl0lIMRnhAXQ15TgG6eezQyb2qVZCO0qEflzZM/V3IsexlOM40JMxVkO56E3F/7xupqK6n1OeZgo4mS+KMmarxJ62YodUAFFsrAkmguq/2mSIBSa6A2nqEtzFk5eJV6teVN3rWqVRL9oooyN0gs6Qi85RA12hFvIQQQ/oCb2gV+PReDbejPf5aMkoMofoD4yPbx1cmZE=</latexit><latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	315	sha1_base64="WNqRVHooK3MRbsx3jlRlp53Ez9s=">AAACB3icbVBNS8NAFNzUrxq/oh49GCyCp5L0Yr0VvHgqVYwttKFsNi/t0s0m7G6EEnr04l/x4kHFq3/Bm//GbRtBWwcWhpk3vH0TpIxK5ThfRmlldW19o7xpbm3v7O5Z+wd3MskEAY8kLBGdAEtglIOnqGLQSQXgOGDQDkaXU799D0LShN+qcQp+jAecRpRgpaW+ddwLYEB5ToArEBPzptk0e8DDH6FvVZyqM4O9TNyCVFCBVt/67IUJyWIdJwxL2	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	316	8DDH6FvVZyqM4O9TNyCVFCBVt/67IUJyWIdJwxL2XWdVPk5FooSBhOzl0lIMRnhAXQ15TgG6eezQyb2qVZCO0qEflzZM/V3IsexlOM40JMxVkO56E3F/7xupqK6n1OeZgo4mS+KMmarxJ62YodUAFFsrAkmguq/2mSIBSa6A2nqEtzFk5eJV6teVN3rWqVRL9oooyN0gs6Qi85RA12hFvIQQQ/oCb2gV+PReDbejPf5aMkoMofoD4yPbx1cmZE=</latexit>RNN	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	317	<latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	318	sha1_base64="WNqRVHooK3MRbsx3jlRlp53Ez9s=">AAACB3icbVBNS8NAFNzUrxq/oh49GCyCp5L0Yr0VvHgqVYwttKFsNi/t0s0m7G6EEnr04l/x4kHFq3/Bm//GbRtBWwcWhpk3vH0TpIxK5ThfRmlldW19o7xpbm3v7O5Z+wd3MskEAY8kLBGdAEtglIOnqGLQSQXgOGDQDkaXU799D0LShN+qcQp+jAecRpRgpaW+ddwLYEB5ToArEBPzptk0e8DDH6FvVZyqM4O9TNyCVFCBVt/67IUJyWIdJwxL2	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	319	8DDH6FvVZyqM4O9TNyCVFCBVt/67IUJyWIdJwxL2XWdVPk5FooSBhOzl0lIMRnhAXQ15TgG6eezQyb2qVZCO0qEflzZM/V3IsexlOM40JMxVkO56E3F/7xupqK6n1OeZgo4mS+KMmarxJ62YodUAFFsrAkmguq/2mSIBSa6A2nqEtzFk5eJV6teVN3rWqVRL9oooyN0gs6Qi85RA12hFvIQQQ/oCb2gV+PReDbejPf5aMkoMofoD4yPbx1cmZE=</latexit><latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	320	sha1_base64="WNqRVHooK3MRbsx3jlRlp53Ez9s=">AAACB3icbVBNS8NAFNzUrxq/oh49GCyCp5L0Yr0VvHgqVYwttKFsNi/t0s0m7G6EEnr04l/x4kHFq3/Bm//GbRtBWwcWhpk3vH0TpIxK5ThfRmlldW19o7xpbm3v7O5Z+wd3MskEAY8kLBGdAEtglIOnqGLQSQXgOGDQDkaXU799D0LShN+qcQp+jAecRpRgpaW+ddwLYEB5ToArEBPzptk0e8DDH6FvVZyqM4O9TNyCVFCBVt/67IUJyWIdJwxL2	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	321	8DDH6FvVZyqM4O9TNyCVFCBVt/67IUJyWIdJwxL2XWdVPk5FooSBhOzl0lIMRnhAXQ15TgG6eezQyb2qVZCO0qEflzZM/V3IsexlOM40JMxVkO56E3F/7xupqK6n1OeZgo4mS+KMmarxJ62YodUAFFsrAkmguq/2mSIBSa6A2nqEtzFk5eJV6teVN3rWqVRL9oooyN0gs6Qi85RA12hFvIQQQ/oCb2gV+PReDbejPf5aMkoMofoD4yPbx1cmZE=</latexit><latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	322	sha1_base64="WNqRVHooK3MRbsx3jlRlp53Ez9s=">AAACB3icbVBNS8NAFNzUrxq/oh49GCyCp5L0Yr0VvHgqVYwttKFsNi/t0s0m7G6EEnr04l/x4kHFq3/Bm//GbRtBWwcWhpk3vH0TpIxK5ThfRmlldW19o7xpbm3v7O5Z+wd3MskEAY8kLBGdAEtglIOnqGLQSQXgOGDQDkaXU799D0LShN+qcQp+jAecRpRgpaW+ddwLYEB5ToArEBPzptk0e8DDH6FvVZyqM4O9TNyCVFCBVt/67IUJyWIdJwxL2	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	323	8DDH6FvVZyqM4O9TNyCVFCBVt/67IUJyWIdJwxL2XWdVPk5FooSBhOzl0lIMRnhAXQ15TgG6eezQyb2qVZCO0qEflzZM/V3IsexlOM40JMxVkO56E3F/7xupqK6n1OeZgo4mS+KMmarxJ62YodUAFFsrAkmguq/2mSIBSa6A2nqEtzFk5eJV6teVN3rWqVRL9oooyN0gs6Qi85RA12hFvIQQQ/oCb2gV+PReDbejPf5aMkoMofoD4yPbx1cmZE=</latexit><latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	324	sha1_base64="WNqRVHooK3MRbsx3jlRlp53Ez9s=">AAACB3icbVBNS8NAFNzUrxq/oh49GCyCp5L0Yr0VvHgqVYwttKFsNi/t0s0m7G6EEnr04l/x4kHFq3/Bm//GbRtBWwcWhpk3vH0TpIxK5ThfRmlldW19o7xpbm3v7O5Z+wd3MskEAY8kLBGdAEtglIOnqGLQSQXgOGDQDkaXU799D0LShN+qcQp+jAecRpRgpaW+ddwLYEB5ToArEBPzptk0e8DDH6FvVZyqM4O9TNyCVFCBVt/67IUJyWIdJwxL2	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	325	8DDH6FvVZyqM4O9TNyCVFCBVt/67IUJyWIdJwxL2XWdVPk5FooSBhOzl0lIMRnhAXQ15TgG6eezQyb2qVZCO0qEflzZM/V3IsexlOM40JMxVkO56E3F/7xupqK6n1OeZgo4mS+KMmarxJ62YodUAFFsrAkmguq/2mSIBSa6A2nqEtzFk5eJV6teVN3rWqVRL9oooyN0gs6Qi85RA12hFvIQQQ/oCb2gV+PReDbejPf5aMkoMofoD4yPbx1cmZE=</latexit>rt1	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	326	<latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	327	sha1_base64="1rbc+gZHaqLmGJLkenaEBa002fg=">AAACDXicbVDLSsNAFJ3UV42vqEs3g22hG0vSjboruHFZwdpCG8pketMOnUzCzEQoIV/gxl9x40LFrXt3/o3Th6CtBwYO59zDnXuChDOlXffLKqytb2xuFbftnd29/QPn8OhOxamk0KIxj2UnIAo4E9DSTHPoJBJIFHBoB+Orqd++B6lYLG71JAE/IkPBQkaJNlLfqfQCGDKRURAaZG6XZT/TZ15etnsgBj9y3ym5NXcGvEq8BSmhBZp957M3iGkam	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	328	15etnsgBj9y3ym5NXcGvEq8BSmhBZp957M3iGkamTjlRKmu5ybaz4jUjHLI7V6qICF0TIbQNVSQCJSfzc7JccUoAxzG0jyh8Uz9nchIpNQkCsxkRPRILXtT8T+vm+rwws+YSFINgs4XhSnHOsbTbvCASaCaTwwhVDLzV0xHRBJqOlC2KcFbPnmVtOq1y5p3Uy81qos2iugEnaIq8tA5aqBr1EQtRNEDekIv6NV6tJ6tN+t9PlqwFplj9AfWxzc1pZuy</latexit><latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	329	sha1_base64="1rbc+gZHaqLmGJLkenaEBa002fg=">AAACDXicbVDLSsNAFJ3UV42vqEs3g22hG0vSjboruHFZwdpCG8pketMOnUzCzEQoIV/gxl9x40LFrXt3/o3Th6CtBwYO59zDnXuChDOlXffLKqytb2xuFbftnd29/QPn8OhOxamk0KIxj2UnIAo4E9DSTHPoJBJIFHBoB+Orqd++B6lYLG71JAE/IkPBQkaJNlLfqfQCGDKRURAaZG6XZT/TZ15etnsgBj9y3ym5NXcGvEq8BSmhBZp957M3iGkam	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	330	15etnsgBj9y3ym5NXcGvEq8BSmhBZp957M3iGkamTjlRKmu5ybaz4jUjHLI7V6qICF0TIbQNVSQCJSfzc7JccUoAxzG0jyh8Uz9nchIpNQkCsxkRPRILXtT8T+vm+rwws+YSFINgs4XhSnHOsbTbvCASaCaTwwhVDLzV0xHRBJqOlC2KcFbPnmVtOq1y5p3Uy81qos2iugEnaIq8tA5aqBr1EQtRNEDekIv6NV6tJ6tN+t9PlqwFplj9AfWxzc1pZuy</latexit><latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	331	sha1_base64="1rbc+gZHaqLmGJLkenaEBa002fg=">AAACDXicbVDLSsNAFJ3UV42vqEs3g22hG0vSjboruHFZwdpCG8pketMOnUzCzEQoIV/gxl9x40LFrXt3/o3Th6CtBwYO59zDnXuChDOlXffLKqytb2xuFbftnd29/QPn8OhOxamk0KIxj2UnIAo4E9DSTHPoJBJIFHBoB+Orqd++B6lYLG71JAE/IkPBQkaJNlLfqfQCGDKRURAaZG6XZT/TZ15etnsgBj9y3ym5NXcGvEq8BSmhBZp957M3iGkam	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	332	15etnsgBj9y3ym5NXcGvEq8BSmhBZp957M3iGkamTjlRKmu5ybaz4jUjHLI7V6qICF0TIbQNVSQCJSfzc7JccUoAxzG0jyh8Uz9nchIpNQkCsxkRPRILXtT8T+vm+rwws+YSFINgs4XhSnHOsbTbvCASaCaTwwhVDLzV0xHRBJqOlC2KcFbPnmVtOq1y5p3Uy81qos2iugEnaIq8tA5aqBr1EQtRNEDekIv6NV6tJ6tN+t9PlqwFplj9AfWxzc1pZuy</latexit><latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	333	sha1_base64="1rbc+gZHaqLmGJLkenaEBa002fg=">AAACDXicbVDLSsNAFJ3UV42vqEs3g22hG0vSjboruHFZwdpCG8pketMOnUzCzEQoIV/gxl9x40LFrXt3/o3Th6CtBwYO59zDnXuChDOlXffLKqytb2xuFbftnd29/QPn8OhOxamk0KIxj2UnIAo4E9DSTHPoJBJIFHBoB+Orqd++B6lYLG71JAE/IkPBQkaJNlLfqfQCGDKRURAaZG6XZT/TZ15etnsgBj9y3ym5NXcGvEq8BSmhBZp957M3iGkam	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	334	15etnsgBj9y3ym5NXcGvEq8BSmhBZp957M3iGkamTjlRKmu5ybaz4jUjHLI7V6qICF0TIbQNVSQCJSfzc7JccUoAxzG0jyh8Uz9nchIpNQkCsxkRPRILXtT8T+vm+rwws+YSFINgs4XhSnHOsbTbvCASaCaTwwhVDLzV0xHRBJqOlC2KcFbPnmVtOq1y5p3Uy81qos2iugEnaIq8tA5aqBr1EQtRNEDekIv6NV6tJ6tN+t9PlqwFplj9AfWxzc1pZuy</latexit>gt1	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	335	<latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	336	sha1_base64="yzX8Dk93KrfNdGHgx6JSbbSoHms=">AAACDXicbVDLSsNAFJ3UV42vqEs3wbbQjSXpRt0V3LisYG2hDWEyuUmHTiZhZiKU0C9w46+4caHi1r07/8bpQ9DWAwOHc+7hzj1BxqhUjvNllNbWNza3ytvmzu7e/oF1eHQn01wQ6JCUpaIXYAmMcugoqhj0MgE4CRh0g9HV1O/eg5A05bdqnIGX4JjTiBKstORbtUEAMeUFAa5ATMxq7BfqzJ1UzQHw8Ef2rYrTcGawV4m7IBW0QNu3PgdhSvJEx	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	337	J1UzQHw8Ef2rYrTcGawV4m7IBW0QNu3PgdhSvJExwnDUvZdJ1NegYWihMHEHOQSMkxGOIa+phwnIL1ids7ErmkltKNU6MeVPVN/JwqcSDlOAj2ZYDWUy95U/M/r5yq68ArKs1wBJ/NFUc5sldrTbuyQCiCKjTXBRFD9V5sMscBEdyBNXYK7fPIq6TQblw33pllp1RdtlNEJOkV15KJz1ELXqI06iKAH9IRe0KvxaDwbb8b7fLRkLDLH6A+Mj28kKJun</latexit><latexit	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]
1806.06157	338	sha1_base64="yzX8Dk93KrfNdGHgx6JSbbSoHms=">AAACDXicbVDLSsNAFJ3UV42vqEs3wbbQjSXpRt0V3LisYG2hDWEyuUmHTiZhZiKU0C9w46+4caHi1r07/8bpQ9DWAwOHc+7hzj1BxqhUjvNllNbWNza3ytvmzu7e/oF1eHQn01wQ6JCUpaIXYAmMcugoqhj0MgE4CRh0g9HV1O/eg5A05bdqnIGX4JjTiBKstORbtUEAMeUFAa5ATMxq7BfqzJ1UzQHw8Ef2rYrTcGawV4m7IBW0QNu3PgdhSvJEx	1806.06157	Object Level Visual Reasoning in Videos	Human activity recognition is typically addressed by detecting key concepts like global and local motion, features related to object classes present in the scene, as well as features related to the global context. The next open challenges in activity recognition require a level of understanding that pushes beyond this and call for models with capabilities for fine distinction and detailed comprehension of interactions between actors and objects in a scene. We propose a model capable of learning to reason about semantically meaningful spatiotemporal interactions in videos. The key to our approach is a choice of performing this reasoning at the object level through the integration of state of the art object detection networks. This allows the model to learn detailed spatial interactions that exist at a semantic, object-interaction relevant level. We evaluate our method on three standard datasets (Twenty-BN Something-Something, VLOG and EPIC Kitchens) and achieve state of the art results on all of them. Finally, we show visualizations of the interactions learned by the model, which illustrate object classes and their interactions corresponding to different activity classes.	http://arxiv.org/pdf/1806.06157	[ "Fabien Baradel", "Natalia Neverova", "Christian Wolf", "Julien Mille", "Greg Mori" ]	[ "cs.CV" ]	Accepted at ECCV 2018 - long version (16 pages + ref)	ECCV 2018	cs.CV	20180616	20180920	[]

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.