Yongxin-Guo
commited on
Commit
•
ee74ab3
1
Parent(s):
9f1b404
Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,59 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: apache-2.0
|
3 |
+
language:
|
4 |
+
- en
|
5 |
+
base_model:
|
6 |
+
- mistralai/Mistral-7B-Instruct-v0.2
|
7 |
+
tags:
|
8 |
+
- video temporal grounding
|
9 |
+
- dense video caption
|
10 |
+
- video highlight detection
|
11 |
+
---
|
12 |
+
|
13 |
+
<h2 align="center"> <a href="https://arxiv.org/abs/2410.05643">TRACE: Temporal Grounding Video LLM via Causal Event Modeling</a></h2>
|
14 |
+
<h5 align="center"> If our project helps you, please give us a star ⭐ on <a href="https://github.com/gyxxyg/TRACE">GitHub</a> and cite our paper!</h2>
|
15 |
+
<h5 align="center">
|
16 |
+
|
17 |
+
## 📰 News
|
18 |
+
|
19 |
+
- **[2024.10.10]** 🔥 Our [code](https://github.com/gyxxyg/TRACE) and [paper](https://arxiv.org/abs/2410.05643) are released!
|
20 |
+
- **[2024.10.10]** 🔥 Our **checkpoints** are available now!
|
21 |
+
|
22 |
+
## Overview
|
23 |
+
|
24 |
+
In this work
|
25 |
+
- We model the videos by a series of events, and propose causal event modeling framework to capture videos' inherent structure.
|
26 |
+
- We present a novel task-interleaved video LLM model, TRACE, tailored to implement the causal event modeling framework through the sequential encoding/decoding of timestamps, salient scores, and textual captions.
|
27 |
+
|
28 |
+
## Model Zoo
|
29 |
+
|
30 |
+
| Checkpoints | Description | URL |
|
31 |
+
| ----------- | ----------- | ----------- |
|
32 |
+
| Initialization | Weights initialized from VideoLLaMA2 | [trace-init](https://huggingface.co/Yongxin-Guo/trace-init) |
|
33 |
+
| Stage-1 | Model checkpoints trained after stage-1 | [trace-stage1](https://huggingface.co/Yongxin-Guo/trace-stage1) |
|
34 |
+
| Stage-2 | Model checkpoints trained after stage-2 | [trace](https://huggingface.co/Yongxin-Guo/trace) |
|
35 |
+
| FT-Charades | Fine-tuned on Charades-STA dataset | [trace-ft-charades](https://huggingface.co/Yongxin-Guo/trace-ft-charades) |
|
36 |
+
| FT-Youcook2 | Fine-tuned on Youcook2 dataset | [trace-ft-youcook2](https://huggingface.co/Yongxin-Guo/trace-ft-youcook2) |
|
37 |
+
| FT-QVHighlights | Fine-tuned on QVHighlights dataset | [trace-ft-qvhighlights](https://huggingface.co/Yongxin-Guo/trace-ft-qvhighlights) |
|
38 |
+
|
39 |
+
#### Results
|
40 |
+
|
41 |
+
| Youcook2 (Zero-Shot) | CIDER | METEOR | SODA_c | F1 |
|
42 |
+
| --- | --- | --- | --- | --- |
|
43 |
+
| TRACE | 8.1 | 2.8 | 2.2 | 22.4 |
|
44 |
+
|
45 |
+
| Charades-STA (Zero-Shot) | 0.3 | 0.5 | 0.7 | mIOU |
|
46 |
+
| --- | --- | --- | --- | --- |
|
47 |
+
| TRACE | 58.6 | 40.3 | 19.4 | 38.7 |
|
48 |
+
|
49 |
+
| QVHighlights (Zero-Shot) | mAP | Hit@1 |
|
50 |
+
| --- | --- | --- |
|
51 |
+
| TRACE | 26.8 | 42.7
|
52 |
+
|
53 |
+
| ActivityNet-DVC | CIDER | METEOR | SODA_c | F1 |
|
54 |
+
| --- | --- | --- | --- | --- |
|
55 |
+
| TRACE | 25.9 | 6.0 | 6.4 | 39.3 |
|
56 |
+
|
57 |
+
| ActivityNet-MR | 0.3 | 0.5 | 0.7 | mIOU |
|
58 |
+
| --- | --- | --- | --- | --- |
|
59 |
+
| TRACE | 53.0 | 37.7 | 24.0 | 39.0 |
|