Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,111 @@
|
|
1 |
-
---
|
2 |
-
license: mit
|
3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: mit
|
3 |
+
task_categories:
|
4 |
+
- visual-question-answering
|
5 |
+
- question-answering
|
6 |
+
language:
|
7 |
+
- en
|
8 |
+
pretty_name: Multi-source Video Captioning
|
9 |
+
size_categories:
|
10 |
+
- 1K<n<10K
|
11 |
+
---
|
12 |
+
|
13 |
+
# Multi-source Video Captioning (MSVC) Dataset Card
|
14 |
+
|
15 |
+
## Dataset details
|
16 |
+
|
17 |
+
**Dataset type:**
|
18 |
+
MSVC is a set of collected video captioning data. It is constructed to ensure a robust and thorough evaluation of Video-LLMs' video-captioning capabilities.
|
19 |
+
|
20 |
+
**Dataset detail:**
|
21 |
+
MSVC is introduced to address limitations in existing video caption benchmarks, MSVC samples 500 videos with human-annotated captions from [MSVD](https://www.aclweb.org/anthology/P11-1020/), [MSRVTT](http://openaccess.thecvf.com/content_cvpr_2016/papers/Xu_MSR-VTT_A_Large_CVPR_2016_paper.pdf), and [VATEX](http://openaccess.thecvf.com/content_ICCV_2019/papers/Wang_VaTeX_A_Large-Scale_High-Quality_Multilingual_Dataset_for_Video-and-Language_Research_ICCV_2019_paper.pdf), ensuring diverse scenarios and domains.
|
22 |
+
Traditional evaluation metrics rely on exact match statistics between generated and ground truth captions, which are limited in capturing the richness of video content. Thus, we use a ChatGPT-assisted evaluation similar to [VideoChatGPT](https://github.com/mbzuai-oryx/Video-ChatGPT/blob/main/quantitative_evaluation/README.md). Both generated and human-annotated captions are evaluated by GPT-3.5-turbo (0613) for Correctness of Information and Detailed Orientation.
|
23 |
+
It is worth noting that each video in the MSVC benchmark is annotated with multiple human-written captions, covering different aspects of the video. This comprehensive annotation ensures a robust and thorough evaluation of Video-LLMs.
|
24 |
+
|
25 |
+
**Data instructions**
|
26 |
+
Please download the raw videos from their official websites and arrange them in the following structure:
|
27 |
+
```bash
|
28 |
+
VideoLLaMA2
|
29 |
+
βββ eval
|
30 |
+
β βββ MSVC
|
31 |
+
| | βββ msvd/
|
32 |
+
| | | βββ lw7pTwpx0K0_38_48.avi
|
33 |
+
| | | βββ ...
|
34 |
+
| | βββ msrvtt/
|
35 |
+
| | | βββ video9921.mp4
|
36 |
+
| | | βββ ...
|
37 |
+
| | βββ vatex/
|
38 |
+
| | | βββ 9giWHf6Pf24.mp4
|
39 |
+
| | | βββ ...
|
40 |
+
```
|
41 |
+
|
42 |
+
|
43 |
+
**GPT3.5 Evaluation Prompt:**
|
44 |
+
```python
|
45 |
+
# Correctness evaluation:
|
46 |
+
{
|
47 |
+
"role": "system",
|
48 |
+
"content":
|
49 |
+
"You are an intelligent chatbot designed for evaluating the factual accuracy of generative outputs for video-based question-answer pairs. "
|
50 |
+
"Your task is to compare the predicted answer with these correct answers and determine if they are factually consistent. Here's how you can accomplish the task:"
|
51 |
+
"------"
|
52 |
+
"##INSTRUCTIONS: "
|
53 |
+
"- Focus on the factual consistency between the predicted answer and the correct answer. The predicted answer should not contain any misinterpretations or misinformation.\n"
|
54 |
+
"- The predicted answer must be factually accurate and align with the video content.\n"
|
55 |
+
"- Consider synonyms or paraphrases as valid matches.\n"
|
56 |
+
"- Evaluate the factual accuracy of the prediction compared to the answer."
|
57 |
+
},
|
58 |
+
{
|
59 |
+
"role": "user",
|
60 |
+
"content":
|
61 |
+
"Please evaluate the following video-based question-answer pair:\n\n"
|
62 |
+
f"Question: {question}\n"
|
63 |
+
f"Correct Answers: {answer}\n"
|
64 |
+
f"Predicted Answer: {pred}\n\n"
|
65 |
+
"Provide your evaluation only as a factual accuracy score where the factual accuracy score is an integer value between 0 and 5, with 5 indicating the highest level of factual consistency. "
|
66 |
+
"Please generate the response in the form of a Python dictionary string with keys 'score', where its value is the factual accuracy score in INTEGER, not STRING."
|
67 |
+
"DO NOT PROVIDE ANY OTHER OUTPUT TEXT OR EXPLANATION. Only provide the Python dictionary string. "
|
68 |
+
"For example, your response should look like this: {''score': 4.8}."
|
69 |
+
}
|
70 |
+
```
|
71 |
+
```python
|
72 |
+
# Detailedness evaluation:
|
73 |
+
{
|
74 |
+
"role": "system",
|
75 |
+
"content": "You are an intelligent chatbot designed for evaluating the detail orientation of generative outputs for video-based question-answer pairs. "
|
76 |
+
"Your task is to compare the predicted answer with these correct answers and determine its level of detail, considering both completeness and specificity. Here's how you can accomplish the task:"
|
77 |
+
"------"
|
78 |
+
"##INSTRUCTIONS: "
|
79 |
+
"- Check if the predicted answer covers all major points from the video. The response should not leave out any key aspects.\n"
|
80 |
+
"- Evaluate whether the predicted answer includes specific details rather than just generic points. It should provide comprehensive information that is tied to specific elements of the video.\n"
|
81 |
+
"- Consider synonyms or paraphrases as valid matches.\n"
|
82 |
+
"- Provide a single evaluation score that reflects the level of detail orientation of the prediction, considering both completeness and specificity.",
|
83 |
+
},
|
84 |
+
{
|
85 |
+
"role": "user",
|
86 |
+
"content": "Please evaluate the following video-based question-answer pair:\n\n"
|
87 |
+
f"Question: {question}\n"
|
88 |
+
f"Correct Answers: {answer}\n"
|
89 |
+
f"Predicted Answer: {pred}\n\n"
|
90 |
+
"Provide your evaluation only as a detail orientation score where the detail orientation score is an integer value between 0 and 5, with 5 indicating the highest level of detail orientation. "
|
91 |
+
"Please generate the response in the form of a Python dictionary string with keys 'score', where its value is the detail orientation score in INTEGER, not STRING."
|
92 |
+
"DO NOT PROVIDE ANY OTHER OUTPUT TEXT OR EXPLANATION. Only provide the Python dictionary string. "
|
93 |
+
"For example, your response should look like this: {''score': 4.8}.",
|
94 |
+
}
|
95 |
+
```
|
96 |
+
|
97 |
+
**Dataset date:**
|
98 |
+
MSVC was released in June 2024.
|
99 |
+
|
100 |
+
**Paper or resources for more information:**
|
101 |
+
https://github.com/DAMO-NLP-SG/VideoLLaMA2
|
102 |
+
|
103 |
+
**Where to send questions or comments about the model:**
|
104 |
+
https://github.com/DAMO-NLP-SG/VideoLLaMA2/issues
|
105 |
+
|
106 |
+
## Intended use
|
107 |
+
**Primary intended uses:**
|
108 |
+
The primary use of MSVC is research on Video-LLMs.
|
109 |
+
|
110 |
+
**Primary intended users:**
|
111 |
+
The primary intended users of the dataset are researchers and hobbyists in computer vision, natural language processing, machine learning, and artificial intelligence.
|