|
|
--- |
|
|
license: mit |
|
|
task_categories: |
|
|
- text-generation |
|
|
language: |
|
|
- en |
|
|
tags: |
|
|
- code |
|
|
pretty_name: OmniACT |
|
|
--- |
|
|
<img src="intro.png" width="700" title="OmniACT"> |
|
|
|
|
|
Dataset for [OmniACT: A Dataset and Benchmark for Enabling Multimodal Generalist Autonomous Agents for Desktop and Web](https://arxiv.org/abs/2402.17553) |
|
|
|
|
|
Splits: |
|
|
|
|
|
| split_name | count | |
|
|
|------------|-------| |
|
|
| train | 6788 | |
|
|
| test | 2020 | |
|
|
| val | 991 | |
|
|
|
|
|
Example datapoint: |
|
|
```json |
|
|
"2849": { |
|
|
"task": "data/tasks/desktop/ibooks/task_1.30.txt", |
|
|
"image": "data/data/desktop/ibooks/screen_1.png", |
|
|
"box": "data/metadata/desktop/boxes/ibooks/screen_1.json" |
|
|
}, |
|
|
``` |
|
|
|
|
|
where: |
|
|
|
|
|
- `task` - contains natural language description ("Task") along with the corresponding PyAutoGUI code ("Output Script"): |
|
|
```text |
|
|
Task: Navigate to see the upcoming titles |
|
|
Output Script: |
|
|
pyautogui.moveTo(1881.5,1116.0) |
|
|
``` |
|
|
- `image` - screen image where the action is performed |
|
|
|
|
|
- `box` - This is the metadata used during evaluation. The json format file contains labels for the interactable elements on the screen and their corresponding bounding boxes. <i>They shouldn't be used while inferencing on the test set.</i> |
|
|
|
|
|
<img src="screen_1.png" width="700" title="example screen image"> |
|
|
|
|
|
|
|
|
To cite OmniACT, please use: |
|
|
``` |
|
|
@misc{kapoor2024omniact, |
|
|
title={OmniACT: A Dataset and Benchmark for Enabling Multimodal Generalist Autonomous Agents for Desktop and Web}, |
|
|
author={Raghav Kapoor and Yash Parag Butala and Melisa Russak and Jing Yu Koh and Kiran Kamble and Waseem Alshikh and Ruslan Salakhutdinov}, |
|
|
year={2024}, |
|
|
eprint={2402.17553}, |
|
|
archivePrefix={arXiv}, |
|
|
primaryClass={cs.AI} |
|
|
} |
|
|
``` |
