Update README.md
Browse files
README.md
CHANGED
|
@@ -65,3 +65,66 @@ configs:
|
|
| 65 |
- split: video_sg_what_action
|
| 66 |
path: data/video_sg_what_action-*
|
| 67 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 65 |
- split: video_sg_what_action
|
| 66 |
path: data/video_sg_what_action-*
|
| 67 |
---
|
| 68 |
+
# Dataset Card for TaskMeAnything-v1-videoqa-2024
|
| 69 |
+
<h2 align="center"> TaskMeAnything-v1-videoqa-2024 benchmark dataset</h2>
|
| 70 |
+
|
| 71 |
+
<h2 align="center"> <a href="https://www.task-me-anything.org/">🌐 Website</a> | <a href="https://arxiv.org/abs/2406.11775">📑 Paper</a> | <a href="https://huggingface.co/collections/jieyuz2/taskmeanything-664ebf028ab2524c0380526a">🤗 Huggingface</a> | <a href="https://huggingface.co/spaces/zixianma/TaskMeAnything-UI">💻 Interface</a></h2>
|
| 72 |
+
|
| 73 |
+
<h5 align="center"> If you like our project, please give us a star ⭐ on GitHub for latest update. </h2>
|
| 74 |
+
|
| 75 |
+
## TaskMeAnything-v1-2024-Videoqa
|
| 76 |
+
[TaskMeAnything-v1-videoqa-2024](https://huggingface.co/datasets/weikaih/TaskMeAnything-v1-videoqa-2024) is a benchmark for reflecting the current progress of MLMs by `automatically` finding tasks that SOTA MLMs struggle with using the TaskMeAnything Top-K queries.
|
| 77 |
+
This benchmark includes 2,394 3d video questions and 1,173 real video questions that the TaskMeAnything algorithm automatically approximated as challenging for over 12 popular MLMs.
|
| 78 |
+
|
| 79 |
+
|
| 80 |
+
The dataset contains 19 splits, while each splits contains 300+ questions from a specific task generator in TaskMeAnything-v1. For each row of dataset, it includes: video, question, options, answer and its corresponding task plan.
|
| 81 |
+
|
| 82 |
+
## Load TaskMeAnything-v1-2024 VideoQA Dataset
|
| 83 |
+
```
|
| 84 |
+
import datasets
|
| 85 |
+
|
| 86 |
+
dataset_name = 'weikaih/TaskMeAnything-v1-videoqa-2024'
|
| 87 |
+
dataset = datasets.load_dataset(dataset_name, split = TASK_GENERATOR_SPLIT)
|
| 88 |
+
```
|
| 89 |
+
where `TASK_GENERATOR_SPLIT` is one of the task generators, eg, `2024_2d_how_many`.
|
| 90 |
+
|
| 91 |
+
## Evaluation Results
|
| 92 |
+
|
| 93 |
+
### Overall
|
| 94 |
+
|
| 95 |
+
|
| 96 |
+
|
| 97 |
+
### Breakdown performance on each task types
|
| 98 |
+
|
| 99 |
+
|
| 100 |
+
|
| 101 |
+
|
| 102 |
+
|
| 103 |
+
|
| 104 |
+
|
| 105 |
+
|
| 106 |
+
|
| 107 |
+
## Out-of-Scope Use
|
| 108 |
+
This dataset should not be used for training models.
|
| 109 |
+
|
| 110 |
+
|
| 111 |
+
## Disclaimers
|
| 112 |
+
**TaskMeAnything** and its associated resources are provided for research and educational purposes only.
|
| 113 |
+
The authors and contributors make no warranties regarding the accuracy or reliability of the data and software.
|
| 114 |
+
Users are responsible for ensuring their use complies with applicable laws and regulations.
|
| 115 |
+
The project is not liable for any damages or losses resulting from the use of these resources.
|
| 116 |
+
|
| 117 |
+
## Contact
|
| 118 |
+
|
| 119 |
+
- Jieyu Zhang: jieyuz2@cs.washington.edu
|
| 120 |
+
|
| 121 |
+
## Citation
|
| 122 |
+
**BibTeX:**
|
| 123 |
+
```bibtex
|
| 124 |
+
@article{zhang2024task,
|
| 125 |
+
title={Task Me Anything},
|
| 126 |
+
author={Zhang, Jieyu and Huang, Weikai and Ma, Zixian and Michel, Oscar and He, Dong and Gupta, Tanmay and Ma, Wei-Chiu and Farhadi, Ali and Kembhavi, Aniruddha and Krishna, Ranjay},
|
| 127 |
+
journal={arXiv preprint arXiv:2406.11775},
|
| 128 |
+
year={2024}
|
| 129 |
+
}
|
| 130 |
+
```
|