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 | 39 | ogTMJZL5XZAJD8TXa/sV/wZ8CoJFqSMFmh0vUknkTRP7TrlROt24GcmKogyjHIYu51cQ0bokPShbakgKeiomCUb43OrJLgnlX3C4Jn6e6MgqdajNLaTKTEDvexNxf+8dm56tahgIssNCDo/1Ms5NhJPa8IJUzY7H1lCqGL2r5gOiCLUdqBdW0KwHHmVhNXKVSW4rZbrtUUbJXSKztAFCtAlqqMb1EAhougJvaA39O48O6/Oh/M5H11zFjsn6A+cyTfnKaEK</latexit>activity | 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 | 40 | head<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 | 41 | sha1_base64="tSmUcjZoatQbRNwVotvgoNbiChk=">AAACEXicbVBNS8NAFNzUrxq/qh69LBZBLyXpxXorePFYwdhCE8pm89ou3WzC7qZQQn+DF/+KFw8qXr1589+4bSNo68DCMPOGt2/ClDOlHefLKq2tb2xulbftnd29/YPK4dG9SjJJwaMJT2QnJAo4E+Bppjl0UgkkDjm0w9H1zG+PQSqWiDs9SSGIyUCwPqNEG6lXufBDGDCRUxAa5NQmVLMx0xM8BBLZPojox+pVqk7NmQOvErcgVVSg1at8+lFCs | 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 | 42 | xM8BBLZPojox+pVqk7NmQOvErcgVVSg1at8+lFCs9jEKSdKdV0n1UFOpGaUw9T2MwUpoSMygK6hgsSggnx+0hSfGSXC/USaJzSeq78TOYmVmsShmYyJHqplbyb+53Uz3W8EORNppkHQxaJ+xrFO8KwfHDEJVPOJIYRKZv6K6ZBI04tp0TYluMsnrxKvXruqubf1arNRtFFGJ+gUnSMXXaImukEt5CGKHtATekGv1qP1bL1Z74vRklVkjtEfWB/feK2eJg==</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 | 43 | sha1_base64="tSmUcjZoatQbRNwVotvgoNbiChk=">AAACEXicbVBNS8NAFNzUrxq/qh69LBZBLyXpxXorePFYwdhCE8pm89ou3WzC7qZQQn+DF/+KFw8qXr1589+4bSNo68DCMPOGt2/ClDOlHefLKq2tb2xulbftnd29/YPK4dG9SjJJwaMJT2QnJAo4E+Bppjl0UgkkDjm0w9H1zG+PQSqWiDs9SSGIyUCwPqNEG6lXufBDGDCRUxAa5NQmVLMx0xM8BBLZPojox+pVqk7NmQOvErcgVVSg1at8+lFCs | 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 | 44 | xM8BBLZPojox+pVqk7NmQOvErcgVVSg1at8+lFCs9jEKSdKdV0n1UFOpGaUw9T2MwUpoSMygK6hgsSggnx+0hSfGSXC/USaJzSeq78TOYmVmsShmYyJHqplbyb+53Uz3W8EORNppkHQxaJ+xrFO8KwfHDEJVPOJIYRKZv6K6ZBI04tp0TYluMsnrxKvXruqubf1arNRtFFGJ+gUnSMXXaImukEt5CGKHtATekGv1qP1bL1Z74vRklVkjtEfWB/feK2eJg==</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 | 45 | sha1_base64="tSmUcjZoatQbRNwVotvgoNbiChk=">AAACEXicbVBNS8NAFNzUrxq/qh69LBZBLyXpxXorePFYwdhCE8pm89ou3WzC7qZQQn+DF/+KFw8qXr1589+4bSNo68DCMPOGt2/ClDOlHefLKq2tb2xulbftnd29/YPK4dG9SjJJwaMJT2QnJAo4E+Bppjl0UgkkDjm0w9H1zG+PQSqWiDs9SSGIyUCwPqNEG6lXufBDGDCRUxAa5NQmVLMx0xM8BBLZPojox+pVqk7NmQOvErcgVVSg1at8+lFCs | 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 | 46 | xM8BBLZPojox+pVqk7NmQOvErcgVVSg1at8+lFCs9jEKSdKdV0n1UFOpGaUw9T2MwUpoSMygK6hgsSggnx+0hSfGSXC/USaJzSeq78TOYmVmsShmYyJHqplbyb+53Uz3W8EORNppkHQxaJ+xrFO8KwfHDEJVPOJIYRKZv6K6ZBI04tp0TYluMsnrxKvXruqubf1arNRtFFGJ+gUnSMXXaImukEt5CGKHtATekGv1qP1bL1Z74vRklVkjtEfWB/feK2eJg==</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 | 47 | sha1_base64="tSmUcjZoatQbRNwVotvgoNbiChk=">AAACEXicbVBNS8NAFNzUrxq/qh69LBZBLyXpxXorePFYwdhCE8pm89ou3WzC7qZQQn+DF/+KFw8qXr1589+4bSNo68DCMPOGt2/ClDOlHefLKq2tb2xulbftnd29/YPK4dG9SjJJwaMJT2QnJAo4E+Bppjl0UgkkDjm0w9H1zG+PQSqWiDs9SSGIyUCwPqNEG6lXufBDGDCRUxAa5NQmVLMx0xM8BBLZPojox+pVqk7NmQOvErcgVVSg1at8+lFCs | 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 | 48 | xM8BBLZPojox+pVqk7NmQOvErcgVVSg1at8+lFCs9jEKSdKdV0n1UFOpGaUw9T2MwUpoSMygK6hgsSggnx+0hSfGSXC/USaJzSeq78TOYmVmsShmYyJHqplbyb+53Uz3W8EORNppkHQxaJ+xrFO8KwfHDEJVPOJIYRKZv6K6ZBI04tp0TYluMsnrxKvXruqubf1arNRtFFGJ+gUnSMXXaImukEt5CGKHtATekGv1qP1bL1Z74vRklVkjtEfWB/feK2eJg==</latexit>object | 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 | 49 | head<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 | 50 | sha1_base64="I0wT7S5QKSPfCmhqWZ9Vmd66HwA=">AAACD3icbVDLSgMxFM3UVx1foy7dBIvoqsx0Y90V3Lis4NhCW0omc9vGZpIhyQhl6Ce48VfcuFBx69adf2P6ELT1QOBwzr3cnBOlnGnj+19OYWV1bX2juOlube/s7nn7B7daZopCSCWXqhkRDZwJCA0zHJqpApJEHBrR8HLiN+5BaSbFjRml0ElIX7Aeo8RYqeudtiPoM5FTEAbU2JXRHVCDB0Bitw0i/jG6Xskv+1PgZRLMSQnNUe96n+1Y0iyx6 | 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 | 51 | 0Bitw0i/jG6Xskv+1PgZRLMSQnNUe96n+1Y0iyx65QTrVuBn5pOTpRhlMPYbWcaUkKHpA8tSwVJQHfyaaAxPrFKjHtS2ScMnqq/N3KSaD1KIjuZEDPQi95E/M9rZaZX7eRMpJkBQWeHehnHRuJJOzhmysbnI0sIVcz+FdMBUYTaDrRrSwgWIy+TsFK+KAfXlVKtOm+jiI7QMTpDATpHNXSF6ihEFD2gJ/SCXp1H59l5c95nowVnvnOI/sD5+AaXkJ0c</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 | 52 | sha1_base64="I0wT7S5QKSPfCmhqWZ9Vmd66HwA=">AAACD3icbVDLSgMxFM3UVx1foy7dBIvoqsx0Y90V3Lis4NhCW0omc9vGZpIhyQhl6Ce48VfcuFBx69adf2P6ELT1QOBwzr3cnBOlnGnj+19OYWV1bX2juOlube/s7nn7B7daZopCSCWXqhkRDZwJCA0zHJqpApJEHBrR8HLiN+5BaSbFjRml0ElIX7Aeo8RYqeudtiPoM5FTEAbU2JXRHVCDB0Bitw0i/jG6Xskv+1PgZRLMSQnNUe96n+1Y0iyx6 | 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 | 53 | 0Bitw0i/jG6Xskv+1PgZRLMSQnNUe96n+1Y0iyx65QTrVuBn5pOTpRhlMPYbWcaUkKHpA8tSwVJQHfyaaAxPrFKjHtS2ScMnqq/N3KSaD1KIjuZEDPQi95E/M9rZaZX7eRMpJkBQWeHehnHRuJJOzhmysbnI0sIVcz+FdMBUYTaDrRrSwgWIy+TsFK+KAfXlVKtOm+jiI7QMTpDATpHNXSF6ihEFD2gJ/SCXp1H59l5c95nowVnvnOI/sD5+AaXkJ0c</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 | 54 | sha1_base64="I0wT7S5QKSPfCmhqWZ9Vmd66HwA=">AAACD3icbVDLSgMxFM3UVx1foy7dBIvoqsx0Y90V3Lis4NhCW0omc9vGZpIhyQhl6Ce48VfcuFBx69adf2P6ELT1QOBwzr3cnBOlnGnj+19OYWV1bX2juOlube/s7nn7B7daZopCSCWXqhkRDZwJCA0zHJqpApJEHBrR8HLiN+5BaSbFjRml0ElIX7Aeo8RYqeudtiPoM5FTEAbU2JXRHVCDB0Bitw0i/jG6Xskv+1PgZRLMSQnNUe96n+1Y0iyx6 | 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 | 55 | 0Bitw0i/jG6Xskv+1PgZRLMSQnNUe96n+1Y0iyx65QTrVuBn5pOTpRhlMPYbWcaUkKHpA8tSwVJQHfyaaAxPrFKjHtS2ScMnqq/N3KSaD1KIjuZEDPQi95E/M9rZaZX7eRMpJkBQWeHehnHRuJJOzhmysbnI0sIVcz+FdMBUYTaDrRrSwgWIy+TsFK+KAfXlVKtOm+jiI7QMTpDATpHNXSF6ihEFD2gJ/SCXp1H59l5c95nowVnvnOI/sD5+AaXkJ0c</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 | 56 | sha1_base64="I0wT7S5QKSPfCmhqWZ9Vmd66HwA=">AAACD3icbVDLSgMxFM3UVx1foy7dBIvoqsx0Y90V3Lis4NhCW0omc9vGZpIhyQhl6Ce48VfcuFBx69adf2P6ELT1QOBwzr3cnBOlnGnj+19OYWV1bX2juOlube/s7nn7B7daZopCSCWXqhkRDZwJCA0zHJqpApJEHBrR8HLiN+5BaSbFjRml0ElIX7Aeo8RYqeudtiPoM5FTEAbU2JXRHVCDB0Bitw0i/jG6Xskv+1PgZRLMSQnNUe96n+1Y0iyx6 | 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 | 57 | 0Bitw0i/jG6Xskv+1PgZRLMSQnNUe96n+1Y0iyx65QTrVuBn5pOTpRhlMPYbWcaUkKHpA8tSwVJQHfyaaAxPrFKjHtS2ScMnqq/N3KSaD1KIjuZEDPQi95E/M9rZaZX7eRMpJkBQWeHehnHRuJJOzhmysbnI0sIVcz+FdMBUYTaDrRrSwgWIy+TsFK+KAfXlVKtOm+jiI7QMTpDATpHNXSF6ihEFD2gJ/SCXp1H59l5c95nowVnvnOI/sD5+AaXkJ0c</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 | 58 | 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 | 59 | <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 | 60 | 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 | 61 | 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 | 62 | 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 | 63 | 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 | 64 | 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 | 65 | 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 | 66 | 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 | 67 | qGlYKIfhjiRccbBDUHEP7GuV/LL/gx4lQQLUkIL1LveRxhLmiV2nHKidTvwU9PJiTKMcpi4YaYhJXRI+tC2VJAEdCefrTfBZ1aJcU8qe4TBM/X3RE4SrcdJZJMJMQO97E3F/7x2ZnrVTs5EmhkQdP5QL+PYSDztCsdMATV8bAmhitm/YjogilDbgXZtCcHyyqukUSlflYPbSqlWXbRRRCfoFJ2jAF2iGrpBddRAFD2gJ/SCXp1H59l5c97n0YKzmDlGf+B8fgM5O6NX</latexit>RoI | 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 | 68 | pool<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 | 69 | sha1_base64="4Rn5dqov7d08Zkstz1UFiuPjd8o=">AAACDHicbVDLSgMxFM34rOOr6tJNsAquykw31l3Bje6qOLbQDiWTuW1DM8mQZIQy9Afc+CtuXKi49QPc+TemD0FbDwQO59zDzT1Rypk2nvflLC2vrK6tFzbcza3tnd3i3v6dlpmiEFDJpWpGRANnAgLDDIdmqoAkEYdGNLgY+417UJpJcWuGKYQJ6QnWZZQYK3WKx+0IekzkFIQBNXJv5BVOpeRuG0T8o3aKJa/sTYAXiT8jJTRDvVP8bMeSZomNU | 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 | 70 | eRuG0T8o3aKJa/sTYAXiT8jJTRDvVP8bMeSZomNU060bvleasKcKMMoh5HbzjSkhA5ID1qWCpKADvPJNSN8YpUYd6WyTxg8UX8ncpJoPUwiO5kQ09fz3lj8z2tlplsNcybSzICg00XdjGMj8bgaHDMF1PChJYQqZv+KaZ8oQm0H2rUl+PMnL5KgUj4v+9eVUq06a6OADtEROkU+OkM1dInqKEAUPaAn9IJenUfn2Xlz3qejS84sc4D+wPn4Bg4mm7k=</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 | 71 | sha1_base64="4Rn5dqov7d08Zkstz1UFiuPjd8o=">AAACDHicbVDLSgMxFM34rOOr6tJNsAquykw31l3Bje6qOLbQDiWTuW1DM8mQZIQy9Afc+CtuXKi49QPc+TemD0FbDwQO59zDzT1Rypk2nvflLC2vrK6tFzbcza3tnd3i3v6dlpmiEFDJpWpGRANnAgLDDIdmqoAkEYdGNLgY+417UJpJcWuGKYQJ6QnWZZQYK3WKx+0IekzkFIQBNXJv5BVOpeRuG0T8o3aKJa/sTYAXiT8jJTRDvVP8bMeSZomNU | 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 | 72 | eRuG0T8o3aKJa/sTYAXiT8jJTRDvVP8bMeSZomNU060bvleasKcKMMoh5HbzjSkhA5ID1qWCpKADvPJNSN8YpUYd6WyTxg8UX8ncpJoPUwiO5kQ09fz3lj8z2tlplsNcybSzICg00XdjGMj8bgaHDMF1PChJYQqZv+KaZ8oQm0H2rUl+PMnL5KgUj4v+9eVUq06a6OADtEROkU+OkM1dInqKEAUPaAn9IJenUfn2Xlz3qejS84sc4D+wPn4Bg4mm7k=</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 | 73 | sha1_base64="4Rn5dqov7d08Zkstz1UFiuPjd8o=">AAACDHicbVDLSgMxFM34rOOr6tJNsAquykw31l3Bje6qOLbQDiWTuW1DM8mQZIQy9Afc+CtuXKi49QPc+TemD0FbDwQO59zDzT1Rypk2nvflLC2vrK6tFzbcza3tnd3i3v6dlpmiEFDJpWpGRANnAgLDDIdmqoAkEYdGNLgY+417UJpJcWuGKYQJ6QnWZZQYK3WKx+0IekzkFIQBNXJv5BVOpeRuG0T8o3aKJa/sTYAXiT8jJTRDvVP8bMeSZomNU | 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 | 74 | eRuG0T8o3aKJa/sTYAXiT8jJTRDvVP8bMeSZomNU060bvleasKcKMMoh5HbzjSkhA5ID1qWCpKADvPJNSN8YpUYd6WyTxg8UX8ncpJoPUwiO5kQ09fz3lj8z2tlplsNcybSzICg00XdjGMj8bgaHDMF1PChJYQqZv+KaZ8oQm0H2rUl+PMnL5KgUj4v+9eVUq06a6OADtEROkU+OkM1dInqKEAUPaAn9IJenUfn2Xlz3qejS84sc4D+wPn4Bg4mm7k=</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 | 75 | sha1_base64="4Rn5dqov7d08Zkstz1UFiuPjd8o=">AAACDHicbVDLSgMxFM34rOOr6tJNsAquykw31l3Bje6qOLbQDiWTuW1DM8mQZIQy9Afc+CtuXKi49QPc+TemD0FbDwQO59zDzT1Rypk2nvflLC2vrK6tFzbcza3tnd3i3v6dlpmiEFDJpWpGRANnAgLDDIdmqoAkEYdGNLgY+417UJpJcWuGKYQJ6QnWZZQYK3WKx+0IekzkFIQBNXJv5BVOpeRuG0T8o3aKJa/sTYAXiT8jJTRDvVP8bMeSZomNU | 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 | 76 | eRuG0T8o3aKJa/sTYAXiT8jJTRDvVP8bMeSZomNU060bvleasKcKMMoh5HbzjSkhA5ID1qWCpKADvPJNSN8YpUYd6WyTxg8UX8ncpJoPUwiO5kQ09fz3lj8z2tlplsNcybSzICg00XdjGMj8bgaHDMF1PChJYQqZv+KaZ8oQm0H2rUl+PMnL5KgUj4v+9eVUq06a6OADtEROkU+OkM1dInqKEAUPaAn9IJenUfn2Xlz3qejS84sc4D+wPn4Bg4mm7k=</latexit>activity | 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 | 77 | loss<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 | 78 | sha1_base64="BcPLS//JtZk28cnHPQ8Mz/Eh5ig=">AAACEXicbVBNS8NAFNzUrxq/oh69LBZBLyXpxXorePFYwdpCG8pm89ou3WzC7qYQQn+DF/+KFw8qXr1589+4bSNo68DCMPOGt2+ChDOlXffLKq2tb2xulbftnd29/QPn8Ohexamk0KIxj2UnIAo4E9DSTHPoJBJIFHBoB+Prmd+egFQsFnc6S8CPyFCwAaNEG6nvXPQCGDKRUxAa5NQmVLMJ0xnmsVJ2D0T4Y/Wdilt158CrxCtIBRVo9p3PXhjTN | 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 | 79 | xnmsVJ2D0T4Y/Wdilt158CrxCtIBRVo9p3PXhjTNDJxyolSXc9NtJ8TqRnlMLV7qYKE0DEZQtdQQSJQfj4/aYrPjBLiQSzNExrP1d+JnERKZVFgJiOiR2rZm4n/ed1UD+p+zkSSahB0sWiQcqxjPOsHh0wC1TwzhFDJzF8xHRFpejEt2qYEb/nkVdKqVa+q3m2t0qgXbZTRCTpF58hDl6iBblATtRBFD+gJvaBX69F6tt6s98VoySoyx+gPrI9vwk+eVQ==</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 | 80 | sha1_base64="BcPLS//JtZk28cnHPQ8Mz/Eh5ig=">AAACEXicbVBNS8NAFNzUrxq/oh69LBZBLyXpxXorePFYwdpCG8pm89ou3WzC7qYQQn+DF/+KFw8qXr1589+4bSNo68DCMPOGt2+ChDOlXffLKq2tb2xulbftnd29/QPn8Ohexamk0KIxj2UnIAo4E9DSTHPoJBJIFHBoB+Prmd+egFQsFnc6S8CPyFCwAaNEG6nvXPQCGDKRUxAa5NQmVLMJ0xnmsVJ2D0T4Y/Wdilt158CrxCtIBRVo9p3PXhjTN | 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 | 81 | xnmsVJ2D0T4Y/Wdilt158CrxCtIBRVo9p3PXhjTNDJxyolSXc9NtJ8TqRnlMLV7qYKE0DEZQtdQQSJQfj4/aYrPjBLiQSzNExrP1d+JnERKZVFgJiOiR2rZm4n/ed1UD+p+zkSSahB0sWiQcqxjPOsHh0wC1TwzhFDJzF8xHRFpejEt2qYEb/nkVdKqVa+q3m2t0qgXbZTRCTpF58hDl6iBblATtRBFD+gJvaBX69F6tt6s98VoySoyx+gPrI9vwk+eVQ==</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 | 82 | sha1_base64="BcPLS//JtZk28cnHPQ8Mz/Eh5ig=">AAACEXicbVBNS8NAFNzUrxq/oh69LBZBLyXpxXorePFYwdpCG8pm89ou3WzC7qYQQn+DF/+KFw8qXr1589+4bSNo68DCMPOGt2+ChDOlXffLKq2tb2xulbftnd29/QPn8Ohexamk0KIxj2UnIAo4E9DSTHPoJBJIFHBoB+Prmd+egFQsFnc6S8CPyFCwAaNEG6nvXPQCGDKRUxAa5NQmVLMJ0xnmsVJ2D0T4Y/Wdilt158CrxCtIBRVo9p3PXhjTN | 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 | 83 | xnmsVJ2D0T4Y/Wdilt158CrxCtIBRVo9p3PXhjTNDJxyolSXc9NtJ8TqRnlMLV7qYKE0DEZQtdQQSJQfj4/aYrPjBLiQSzNExrP1d+JnERKZVFgJiOiR2rZm4n/ed1UD+p+zkSSahB0sWiQcqxjPOsHh0wC1TwzhFDJzF8xHRFpejEt2qYEb/nkVdKqVa+q3m2t0qgXbZTRCTpF58hDl6iBblATtRBFD+gJvaBX69F6tt6s98VoySoyx+gPrI9vwk+eVQ==</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 | 84 | sha1_base64="BcPLS//JtZk28cnHPQ8Mz/Eh5ig=">AAACEXicbVBNS8NAFNzUrxq/oh69LBZBLyXpxXorePFYwdpCG8pm89ou3WzC7qYQQn+DF/+KFw8qXr1589+4bSNo68DCMPOGt2+ChDOlXffLKq2tb2xulbftnd29/QPn8Ohexamk0KIxj2UnIAo4E9DSTHPoJBJIFHBoB+Prmd+egFQsFnc6S8CPyFCwAaNEG6nvXPQCGDKRUxAa5NQmVLMJ0xnmsVJ2D0T4Y/Wdilt158CrxCtIBRVo9p3PXhjTN | 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 | 85 | xnmsVJ2D0T4Y/Wdilt158CrxCtIBRVo9p3PXhjTNDJxyolSXc9NtJ8TqRnlMLV7qYKE0DEZQtdQQSJQfj4/aYrPjBLiQSzNExrP1d+JnERKZVFgJiOiR2rZm4n/ed1UD+p+zkSSahB0sWiQcqxjPOsHh0wC1TwzhFDJzF8xHRFpejEt2qYEb/nkVdKqVa+q3m2t0qgXbZTRCTpF58hDl6iBblATtRBFD+gJvaBX69F6tt6s98VoySoyx+gPrI9vwk+eVQ==</latexit>object | 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 | 86 | class loss | 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 | 87 | <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 | 88 | sha1_base64="52a627od74BeVjKBj74VpiEI2Rg=">AAACFXicbVBNSwMxEM3Wr7p+rXr0EiyCF8tuL9ZbwYvHCq4ttKVk02kbm02WJCuUpb/Ci3/FiwcVr4I3/41pu4K2Pgg83puZzLwo4Uwb3/9yCiura+sbxU13a3tnd8/bP7jVMlUUQiq5VM2IaOBMQGiY4dBMFJA44tCIRpdTv3EPSjMpbsw4gU5MBoL1GSXGSl3vrB3BgImMgjCgJq6M7oAaTDnRGnOptdsG0fuxu17JL/sz4GUS5KSEctS73me7J | 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 | 89 | DnRGnOptdsG0fuxu17JL/sz4GUS5KSEctS73me7J2ka2/bZyFbgJ6aTEWUY5TBx26mGhNARGUDLUkFi0J1sdtYEn1ilh/tS2SfsTlP1d0dGYq3HcWQrY2KGetGbiv95rdT0q52MiSQ1IOj8o37KsZF4mhHuMWVD4GNLCFXM7orpkChCbQbatSEEiycvk7BSvigH15VSrZqnUURH6BidogCdoxq6QnUUIooe0BN6Qa/Oo/PsvDnv89KCk/ccoj9wPr4Bd6SfvQ==</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 | 90 | sha1_base64="52a627od74BeVjKBj74VpiEI2Rg=">AAACFXicbVBNSwMxEM3Wr7p+rXr0EiyCF8tuL9ZbwYvHCq4ttKVk02kbm02WJCuUpb/Ci3/FiwcVr4I3/41pu4K2Pgg83puZzLwo4Uwb3/9yCiura+sbxU13a3tnd8/bP7jVMlUUQiq5VM2IaOBMQGiY4dBMFJA44tCIRpdTv3EPSjMpbsw4gU5MBoL1GSXGSl3vrB3BgImMgjCgJq6M7oAaTDnRGnOptdsG0fuxu17JL/sz4GUS5KSEctS73me7J | 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 | 91 | DnRGnOptdsG0fuxu17JL/sz4GUS5KSEctS73me7J2ka2/bZyFbgJ6aTEWUY5TBx26mGhNARGUDLUkFi0J1sdtYEn1ilh/tS2SfsTlP1d0dGYq3HcWQrY2KGetGbiv95rdT0q52MiSQ1IOj8o37KsZF4mhHuMWVD4GNLCFXM7orpkChCbQbatSEEiycvk7BSvigH15VSrZqnUURH6BidogCdoxq6QnUUIooe0BN6Qa/Oo/PsvDnv89KCk/ccoj9wPr4Bd6SfvQ==</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 | 92 | sha1_base64="52a627od74BeVjKBj74VpiEI2Rg=">AAACFXicbVBNSwMxEM3Wr7p+rXr0EiyCF8tuL9ZbwYvHCq4ttKVk02kbm02WJCuUpb/Ci3/FiwcVr4I3/41pu4K2Pgg83puZzLwo4Uwb3/9yCiura+sbxU13a3tnd8/bP7jVMlUUQiq5VM2IaOBMQGiY4dBMFJA44tCIRpdTv3EPSjMpbsw4gU5MBoL1GSXGSl3vrB3BgImMgjCgJq6M7oAaTDnRGnOptdsG0fuxu17JL/sz4GUS5KSEctS73me7J | 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 | 93 | DnRGnOptdsG0fuxu17JL/sz4GUS5KSEctS73me7J2ka2/bZyFbgJ6aTEWUY5TBx26mGhNARGUDLUkFi0J1sdtYEn1ilh/tS2SfsTlP1d0dGYq3HcWQrY2KGetGbiv95rdT0q52MiSQ1IOj8o37KsZF4mhHuMWVD4GNLCFXM7orpkChCbQbatSEEiycvk7BSvigH15VSrZqnUURH6BidogCdoxq6QnUUIooe0BN6Qa/Oo/PsvDnv89KCk/ccoj9wPr4Bd6SfvQ==</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 | 94 | sha1_base64="52a627od74BeVjKBj74VpiEI2Rg=">AAACFXicbVBNSwMxEM3Wr7p+rXr0EiyCF8tuL9ZbwYvHCq4ttKVk02kbm02WJCuUpb/Ci3/FiwcVr4I3/41pu4K2Pgg83puZzLwo4Uwb3/9yCiura+sbxU13a3tnd8/bP7jVMlUUQiq5VM2IaOBMQGiY4dBMFJA44tCIRpdTv3EPSjMpbsw4gU5MBoL1GSXGSl3vrB3BgImMgjCgJq6M7oAaTDnRGnOptdsG0fuxu17JL/sz4GUS5KSEctS73me7J | 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 | 95 | DnRGnOptdsG0fuxu17JL/sz4GUS5KSEctS73me7J2ka2/bZyFbgJ6aTEWUY5TBx26mGhNARGUDLUkFi0J1sdtYEn1ilh/tS2SfsTlP1d0dGYq3HcWQrY2KGetGbiv95rdT0q52MiSQ1IOj8o37KsZF4mhHuMWVD4GNLCFXM7orpkChCbQbatSEEiycvk7BSvigH15VSrZqnUURH6BidogCdoxq6QnUUIooe0BN6Qa/Oo/PsvDnv89KCk/ccoj9wPr4Bd6SfvQ==</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 | 96 | pairwisetemporal sampling | 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 | 97 | <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 | 98 | sha1_base64="IMkmLAb5bdMJnxwm+q7K1alA/fA=">AAACIHicbVBNSwMxFMz6WdevqkcvwSJ4Kru9WG8FLx4ruFrolvI2fa2hSXZJskpZ+le8+Fe8eFDRm/4a01pBqwOBYWYeeW+STHBjg+DdW1hcWl5ZLa356xubW9vlnd1Lk+aaYcRSkepWAgYFVxhZbgW2Mo0gE4FXyfB04l/doDY8VRd2lGFHwkDxPmdgndQt1+MEB1wVDJVFPfYz4PqWG4xjalFmqQZBDUi3ihr4Mared7JbrgTVYAr6l4QzUiEzN | 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 | 99 | lFmqQZBDUi3ihr4Mared7JbrgTVYAr6l4QzUiEzNLvlt7iXsly6cSbAmHYYZLZTgLacCRz7cW4wAzaEAbYdVSDRdIrphWN66JQe7afaPWXpVP05UYA0ZiQTl5Rgr828NxH/89q57dc7BVdZblGxr4/6uaA2pZO6aI9rZFaMHAGmuduVsmvQwFwHxnclhPMn/yVRrXpSDc9rlUZ91kaJ7JMDckRCckwa5Iw0SUQYuSMP5Ik8e/feo/fivX5FF7zZzB75Be/jE9LypLg=</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 | 100 | sha1_base64="IMkmLAb5bdMJnxwm+q7K1alA/fA=">AAACIHicbVBNSwMxFMz6WdevqkcvwSJ4Kru9WG8FLx4ruFrolvI2fa2hSXZJskpZ+le8+Fe8eFDRm/4a01pBqwOBYWYeeW+STHBjg+DdW1hcWl5ZLa356xubW9vlnd1Lk+aaYcRSkepWAgYFVxhZbgW2Mo0gE4FXyfB04l/doDY8VRd2lGFHwkDxPmdgndQt1+MEB1wVDJVFPfYz4PqWG4xjalFmqQZBDUi3ihr4Mared7JbrgTVYAr6l4QzUiEzN | 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 | 101 | lFmqQZBDUi3ihr4Mared7JbrgTVYAr6l4QzUiEzNLvlt7iXsly6cSbAmHYYZLZTgLacCRz7cW4wAzaEAbYdVSDRdIrphWN66JQe7afaPWXpVP05UYA0ZiQTl5Rgr828NxH/89q57dc7BVdZblGxr4/6uaA2pZO6aI9rZFaMHAGmuduVsmvQwFwHxnclhPMn/yVRrXpSDc9rlUZ91kaJ7JMDckRCckwa5Iw0SUQYuSMP5Ik8e/feo/fivX5FF7zZzB75Be/jE9LypLg=</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 | 102 | sha1_base64="IMkmLAb5bdMJnxwm+q7K1alA/fA=">AAACIHicbVBNSwMxFMz6WdevqkcvwSJ4Kru9WG8FLx4ruFrolvI2fa2hSXZJskpZ+le8+Fe8eFDRm/4a01pBqwOBYWYeeW+STHBjg+DdW1hcWl5ZLa356xubW9vlnd1Lk+aaYcRSkepWAgYFVxhZbgW2Mo0gE4FXyfB04l/doDY8VRd2lGFHwkDxPmdgndQt1+MEB1wVDJVFPfYz4PqWG4xjalFmqQZBDUi3ihr4Mared7JbrgTVYAr6l4QzUiEzN | 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 | 103 | lFmqQZBDUi3ihr4Mared7JbrgTVYAr6l4QzUiEzNLvlt7iXsly6cSbAmHYYZLZTgLacCRz7cW4wAzaEAbYdVSDRdIrphWN66JQe7afaPWXpVP05UYA0ZiQTl5Rgr828NxH/89q57dc7BVdZblGxr4/6uaA2pZO6aI9rZFaMHAGmuduVsmvQwFwHxnclhPMn/yVRrXpSDc9rlUZ91kaJ7JMDckRCckwa5Iw0SUQYuSMP5Ik8e/feo/fivX5FF7zZzB75Be/jE9LypLg=</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 | 104 | sha1_base64="IMkmLAb5bdMJnxwm+q7K1alA/fA=">AAACIHicbVBNSwMxFMz6WdevqkcvwSJ4Kru9WG8FLx4ruFrolvI2fa2hSXZJskpZ+le8+Fe8eFDRm/4a01pBqwOBYWYeeW+STHBjg+DdW1hcWl5ZLa356xubW9vlnd1Lk+aaYcRSkepWAgYFVxhZbgW2Mo0gE4FXyfB04l/doDY8VRd2lGFHwkDxPmdgndQt1+MEB1wVDJVFPfYz4PqWG4xjalFmqQZBDUi3ihr4Mared7JbrgTVYAr6l4QzUiEzN | 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 | 105 | lFmqQZBDUi3ihr4Mared7JbrgTVYAr6l4QzUiEzNLvlt7iXsly6cSbAmHYYZLZTgLacCRz7cW4wAzaEAbYdVSDRdIrphWN66JQe7afaPWXpVP05UYA0ZiQTl5Rgr828NxH/89q57dc7BVdZblGxr4/6uaA2pZO6aI9rZFaMHAGmuduVsmvQwFwHxnclhPMn/yVRrXpSDc9rlUZ91kaJ7JMDckRCckwa5Iw0SUQYuSMP5Ik8e/feo/fivX5FF7zZzB75Be/jE9LypLg=</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 | 106 | activity features<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 | 107 | sha1_base64="r+7eRHfo+xMX1EIiEEcQFlIGw6w=">AAACFXicbVC7SgNBFJ2Nr7i+opY2g0GwMeymMXYBG8sIxgSSEGYnd5Mhs7PLzN1AWPIVNv6KjYWKrWDn3zh5CJp4YOBwzj3cuSdIpDDoeV9Obm19Y3Mrv+3u7O7tHxQOj+5NnGoOdR7LWDcDZkAKBXUUKKGZaGBRIKERDK+nfmME2ohY3eE4gU7E+kqEgjO0Urdw0Q6gL1TGQSHoics4ipHAMQ2BYarBuG1QvR+7Wyh6JW8Gukr8BSmSBWrdwme7F | 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 | 108 | Q2BYarBuG1QvR+7Wyh6JW8Gukr8BSmSBWrdwme7F/M0snEumTEt30uwkzGNgkuYuO3UQML4kPWhZaliEZhONjtrQs+s0qNhrO1TSGfq70TGImPGUWAnI4YDs+xNxf+8VophpZMJlaQIis8XhamkGNNpR7QnNHCUY0sY18L+lfIB07Yb26RrS/CXT14l9XLpquTflovVyqKNPDkhp+Sc+OSSVMkNqZE64eSBPJEX8uo8Os/Om/M+H805i8wx+QPn4xsNc6Ab</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 | 109 | sha1_base64="r+7eRHfo+xMX1EIiEEcQFlIGw6w=">AAACFXicbVC7SgNBFJ2Nr7i+opY2g0GwMeymMXYBG8sIxgSSEGYnd5Mhs7PLzN1AWPIVNv6KjYWKrWDn3zh5CJp4YOBwzj3cuSdIpDDoeV9Obm19Y3Mrv+3u7O7tHxQOj+5NnGoOdR7LWDcDZkAKBXUUKKGZaGBRIKERDK+nfmME2ohY3eE4gU7E+kqEgjO0Urdw0Q6gL1TGQSHoics4ipHAMQ2BYarBuG1QvR+7Wyh6JW8Gukr8BSmSBWrdwme7F | 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 | 110 | Q2BYarBuG1QvR+7Wyh6JW8Gukr8BSmSBWrdwme7F/M0snEumTEt30uwkzGNgkuYuO3UQML4kPWhZaliEZhONjtrQs+s0qNhrO1TSGfq70TGImPGUWAnI4YDs+xNxf+8VophpZMJlaQIis8XhamkGNNpR7QnNHCUY0sY18L+lfIB07Yb26RrS/CXT14l9XLpquTflovVyqKNPDkhp+Sc+OSSVMkNqZE64eSBPJEX8uo8Os/Om/M+H805i8wx+QPn4xsNc6Ab</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 | 111 | sha1_base64="r+7eRHfo+xMX1EIiEEcQFlIGw6w=">AAACFXicbVC7SgNBFJ2Nr7i+opY2g0GwMeymMXYBG8sIxgSSEGYnd5Mhs7PLzN1AWPIVNv6KjYWKrWDn3zh5CJp4YOBwzj3cuSdIpDDoeV9Obm19Y3Mrv+3u7O7tHxQOj+5NnGoOdR7LWDcDZkAKBXUUKKGZaGBRIKERDK+nfmME2ohY3eE4gU7E+kqEgjO0Urdw0Q6gL1TGQSHoics4ipHAMQ2BYarBuG1QvR+7Wyh6JW8Gukr8BSmSBWrdwme7F | 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 | 112 | Q2BYarBuG1QvR+7Wyh6JW8Gukr8BSmSBWrdwme7F/M0snEumTEt30uwkzGNgkuYuO3UQML4kPWhZaliEZhONjtrQs+s0qNhrO1TSGfq70TGImPGUWAnI4YDs+xNxf+8VophpZMJlaQIis8XhamkGNNpR7QnNHCUY0sY18L+lfIB07Yb26RrS/CXT14l9XLpquTflovVyqKNPDkhp+Sc+OSSVMkNqZE64eSBPJEX8uo8Os/Om/M+H805i8wx+QPn4xsNc6Ab</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 | 113 | sha1_base64="r+7eRHfo+xMX1EIiEEcQFlIGw6w=">AAACFXicbVC7SgNBFJ2Nr7i+opY2g0GwMeymMXYBG8sIxgSSEGYnd5Mhs7PLzN1AWPIVNv6KjYWKrWDn3zh5CJp4YOBwzj3cuSdIpDDoeV9Obm19Y3Mrv+3u7O7tHxQOj+5NnGoOdR7LWDcDZkAKBXUUKKGZaGBRIKERDK+nfmME2ohY3eE4gU7E+kqEgjO0Urdw0Q6gL1TGQSHoics4ipHAMQ2BYarBuG1QvR+7Wyh6JW8Gukr8BSmSBWrdwme7F | 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 | 114 | Q2BYarBuG1QvR+7Wyh6JW8Gukr8BSmSBWrdwme7F/M0snEumTEt30uwkzGNgkuYuO3UQML4kPWhZaliEZhONjtrQs+s0qNhrO1TSGfq70TGImPGUWAnI4YDs+xNxf+8VophpZMJlaQIis8XhamkGNNpR7QnNHCUY0sY18L+lfIB07Yb26RrS/CXT14l9XLpquTflovVyqKNPDkhp+Sc+OSSVMkNqZE64eSBPJEX8uo8Os/Om/M+H805i8wx+QPn4xsNc6Ab</latexit>n⇥2Dsets | 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 | 115 | ofobject features | 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 | 116 | <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 | 117 | sha1_base64="viP+YHsQOd91KFvZysvFMJg/WIY=">AAACK3icbVBNSwMxFMz67fpV9egl2Aqeym4v6k3Qg8cKVgvdUrLp2xrNJkvyVihLf5AX/4ogHqx49X+Y1graOhAYZt6Q9ybOpLAYBENvbn5hcWl5ZdVfW9/Y3Cpt71xbnRsODa6lNs2YWZBCQQMFSmhmBlgaS7iJ789G/s0DGCu0usJ+Bu2U9ZRIBGfopE7pLIqhJ1TBQSGYgV9RRYQiBTuonVeoBbRUJ1Hk6/gOONIEGOYGrB+B6v6EOqVyUA3Go | 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 | 118 | VeoBbRUJ1Hk6/gOONIEGOYGrB+B6v6EOqVyUA3GoLMknJAymaDeKb1EXc3z1MW5ZNa2wiDDdsEMCi5h4Ee5hYzxe9aDlqOKuX3axfjYAT1wSpcm2rinkI7V34mCpdb209hNpgxv7bQ3Ev/zWjkmx+1CqCxHUPz7oySXFDUdNUe7wrgKZN8Rxo1wu1J+ywzjrgPruxLC6ZNnSaNWPamGl7Xy6fGkjRWyR/bJIQnJETklF6ROGoSTR/JM3sjQe/JevXfv43t0zptkdskfeJ9f7xuoOQ==</latexit><latexi | 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 | 119 | 3t0zptkdskfeJ9f7xuoOQ==</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 | 120 | sha1_base64="viP+YHsQOd91KFvZysvFMJg/WIY=">AAACK3icbVBNSwMxFMz67fpV9egl2Aqeym4v6k3Qg8cKVgvdUrLp2xrNJkvyVihLf5AX/4ogHqx49X+Y1graOhAYZt6Q9ybOpLAYBENvbn5hcWl5ZdVfW9/Y3Cpt71xbnRsODa6lNs2YWZBCQQMFSmhmBlgaS7iJ789G/s0DGCu0usJ+Bu2U9ZRIBGfopE7pLIqhJ1TBQSGYgV9RRYQiBTuonVeoBbRUJ1Hk6/gOONIEGOYGrB+B6v6EOqVyUA3Go | 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 | 121 | VeoBbRUJ1Hk6/gOONIEGOYGrB+B6v6EOqVyUA3GoLMknJAymaDeKb1EXc3z1MW5ZNa2wiDDdsEMCi5h4Ee5hYzxe9aDlqOKuX3axfjYAT1wSpcm2rinkI7V34mCpdb209hNpgxv7bQ3Ev/zWjkmx+1CqCxHUPz7oySXFDUdNUe7wrgKZN8Rxo1wu1J+ywzjrgPruxLC6ZNnSaNWPamGl7Xy6fGkjRWyR/bJIQnJETklF6ROGoSTR/JM3sjQe/JevXfv43t0zptkdskfeJ9f7xuoOQ==</latexit><latexi | 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 | 122 | 3t0zptkdskfeJ9f7xuoOQ==</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 | 123 | sha1_base64="viP+YHsQOd91KFvZysvFMJg/WIY=">AAACK3icbVBNSwMxFMz67fpV9egl2Aqeym4v6k3Qg8cKVgvdUrLp2xrNJkvyVihLf5AX/4ogHqx49X+Y1graOhAYZt6Q9ybOpLAYBENvbn5hcWl5ZdVfW9/Y3Cpt71xbnRsODa6lNs2YWZBCQQMFSmhmBlgaS7iJ789G/s0DGCu0usJ+Bu2U9ZRIBGfopE7pLIqhJ1TBQSGYgV9RRYQiBTuonVeoBbRUJ1Hk6/gOONIEGOYGrB+B6v6EOqVyUA3Go | 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 | 124 | VeoBbRUJ1Hk6/gOONIEGOYGrB+B6v6EOqVyUA3GoLMknJAymaDeKb1EXc3z1MW5ZNa2wiDDdsEMCi5h4Ee5hYzxe9aDlqOKuX3axfjYAT1wSpcm2rinkI7V34mCpdb209hNpgxv7bQ3Ev/zWjkmx+1CqCxHUPz7oySXFDUdNUe7wrgKZN8Rxo1wu1J+ywzjrgPruxLC6ZNnSaNWPamGl7Xy6fGkjRWyR/bJIQnJETklF6ROGoSTR/JM3sjQe/JevXfv43t0zptkdskfeJ9f7xuoOQ==</latexit><latexi | 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 | 125 | 3t0zptkdskfeJ9f7xuoOQ==</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 | 126 | sha1_base64="viP+YHsQOd91KFvZysvFMJg/WIY=">AAACK3icbVBNSwMxFMz67fpV9egl2Aqeym4v6k3Qg8cKVgvdUrLp2xrNJkvyVihLf5AX/4ogHqx49X+Y1graOhAYZt6Q9ybOpLAYBENvbn5hcWl5ZdVfW9/Y3Cpt71xbnRsODa6lNs2YWZBCQQMFSmhmBlgaS7iJ789G/s0DGCu0usJ+Bu2U9ZRIBGfopE7pLIqhJ1TBQSGYgV9RRYQiBTuonVeoBbRUJ1Hk6/gOONIEGOYGrB+B6v6EOqVyUA3Go | 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 | 127 | VeoBbRUJ1Hk6/gOONIEGOYGrB+B6v6EOqVyUA3GoLMknJAymaDeKb1EXc3z1MW5ZNa2wiDDdsEMCi5h4Ee5hYzxe9aDlqOKuX3axfjYAT1wSpcm2rinkI7V34mCpdb209hNpgxv7bQ3Ev/zWjkmx+1CqCxHUPz7oySXFDUdNUe7wrgKZN8Rxo1wu1J+ywzjrgPruxLC6ZNnSaNWPamGl7Xy6fGkjRWyR/bJIQnJETklF6ROGoSTR/JM3sjQe/JevXfv43t0zptkdskfeJ9f7xuoOQ==</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 | 128 | n⇥2Dsets ofobject masks | 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 | 129 | <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 | 130 | sha1_base64="wsaTEZjazwjgqpKPonLLv8b2h38=">AAACKHicbVDLSgMxFM3UVx1fVZdugq3gqsx0Y90VdOGygrWFTimZ9E4bm0mGJCOUob/jxl9xo6DSrV9i+hC09UDgcM495N4TJpxp43kTJ7e2vrG5ld92d3b39g8Kh0f3WqaKQoNKLlUrJBo4E9AwzHBoJQpIHHJohsOrqd98BKWZFHdmlEAnJn3BIkaJsVK3UAtC6DORURAG1NgtiSwwLAY9rlyXsAajsYyCwJXhA1CDY6KH2g1A9H4S3ULRK3sz4 | 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 | 131 | lyXsAajsYyCwJXhA1CDY6KH2g1A9H4S3ULRK3sz4FXiL0gRLVDvFt6CnqRpbOOUE63bvpeYTkaUYZTD2A1SDQmhQ9KHtqWC2GU62ezSMT6zSg9HUtknDJ6pvxMZibUexaGdjIkZ6GVvKv7ntVMTVTsZE0lqQND5R1HKsZF4WhvuMWXv5yNLCFXM7orpgChCbQfatSX4yyevkkalfFn2byvFWnXRRh6doFN0jnx0gWroBtVRA1H0hF7QO/pwnp1X59OZzEdzziJzjP7A+foGUQGm2w==</latexit><latexi | 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 | 132 | EdzziJzjP7A+foGUQGm2w==</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 | 133 | sha1_base64="wsaTEZjazwjgqpKPonLLv8b2h38=">AAACKHicbVDLSgMxFM3UVx1fVZdugq3gqsx0Y90VdOGygrWFTimZ9E4bm0mGJCOUob/jxl9xo6DSrV9i+hC09UDgcM495N4TJpxp43kTJ7e2vrG5ld92d3b39g8Kh0f3WqaKQoNKLlUrJBo4E9AwzHBoJQpIHHJohsOrqd98BKWZFHdmlEAnJn3BIkaJsVK3UAtC6DORURAG1NgtiSwwLAY9rlyXsAajsYyCwJXhA1CDY6KH2g1A9H4S3ULRK3sz4 | 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 | 134 | lyXsAajsYyCwJXhA1CDY6KH2g1A9H4S3ULRK3sz4FXiL0gRLVDvFt6CnqRpbOOUE63bvpeYTkaUYZTD2A1SDQmhQ9KHtqWC2GU62ezSMT6zSg9HUtknDJ6pvxMZibUexaGdjIkZ6GVvKv7ntVMTVTsZE0lqQND5R1HKsZF4WhvuMWXv5yNLCFXM7orpgChCbQfatSX4yyevkkalfFn2byvFWnXRRh6doFN0jnx0gWroBtVRA1H0hF7QO/pwnp1X59OZzEdzziJzjP7A+foGUQGm2w==</latexit><latexi | 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 | 135 | EdzziJzjP7A+foGUQGm2w==</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 | 136 | sha1_base64="wsaTEZjazwjgqpKPonLLv8b2h38=">AAACKHicbVDLSgMxFM3UVx1fVZdugq3gqsx0Y90VdOGygrWFTimZ9E4bm0mGJCOUob/jxl9xo6DSrV9i+hC09UDgcM495N4TJpxp43kTJ7e2vrG5ld92d3b39g8Kh0f3WqaKQoNKLlUrJBo4E9AwzHBoJQpIHHJohsOrqd98BKWZFHdmlEAnJn3BIkaJsVK3UAtC6DORURAG1NgtiSwwLAY9rlyXsAajsYyCwJXhA1CDY6KH2g1A9H4S3ULRK3sz4 | 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 | 137 | lyXsAajsYyCwJXhA1CDY6KH2g1A9H4S3ULRK3sz4FXiL0gRLVDvFt6CnqRpbOOUE63bvpeYTkaUYZTD2A1SDQmhQ9KHtqWC2GU62ezSMT6zSg9HUtknDJ6pvxMZibUexaGdjIkZ6GVvKv7ntVMTVTsZE0lqQND5R1HKsZF4WhvuMWXv5yNLCFXM7orpgChCbQfatSX4yyevkkalfFn2byvFWnXRRh6doFN0jnx0gWroBtVRA1H0hF7QO/pwnp1X59OZzEdzziJzjP7A+foGUQGm2w==</latexit><latexi | 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 | 138 | EdzziJzjP7A+foGUQGm2w==</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 | [] |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.