Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,38 @@
|
|
1 |
-
---
|
2 |
-
license: apache-2.0
|
3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: apache-2.0
|
3 |
+
task_categories:
|
4 |
+
- video-classification
|
5 |
+
language:
|
6 |
+
- en
|
7 |
+
size_categories:
|
8 |
+
- 1B<n<10B
|
9 |
+
---
|
10 |
+
|
11 |
+
## 📙 Overview
|
12 |
+
|
13 |
+
The metadata for Ego4d training set, with paired howto100m video clips. The ego-exo pair is constructed by choosing the ones with shared nouns/verbs.
|
14 |
+
|
15 |
+
Each sample represents a short video clip, which consists of
|
16 |
+
- vid: the initial video id.
|
17 |
+
- start_second: the start timestamp of the narration.
|
18 |
+
- end_second: the end timestamp of the narration.
|
19 |
+
- text: the original narration.
|
20 |
+
- noun: a list containing the index of nouns in the Ego4d noun vocabulary.
|
21 |
+
- verb: a list containing the index of verbs in the Ego4d verb vocabulary.
|
22 |
+
- nv_index: a list of numbers indexing the [paired howto100m clips](https://huggingface.co/datasets/Jazzcharles/HowTo100M_llama3_refined_caption).
|
23 |
+
|
24 |
+
The noun and verb vocabularies can be found in EgoVLP (https://drive.google.com/drive/folders/16fUv5rrZmt06Ty3QAEweDpveC-84RI9Z).
|
25 |
+
|
26 |
+
## 🏋️ How-To-Use
|
27 |
+
|
28 |
+
Please refer to code [EgoInstructor](https://github.com/Jazzcharles/Egoinstructor/) for details.
|
29 |
+
|
30 |
+
## 🎓 Citation
|
31 |
+
```
|
32 |
+
@article{xu2024retrieval,
|
33 |
+
title={Retrieval-augmented egocentric video captioning},
|
34 |
+
author={Xu, Jilan and Huang, Yifei and Hou, Junlin and Chen, Guo and Zhang, Yuejie and Feng, Rui and Xie, Weidi},
|
35 |
+
journal={arXiv preprint arXiv:2401.00789},
|
36 |
+
year={2024}
|
37 |
+
}
|
38 |
+
`
|