| --- |
| license: mit |
| task_categories: |
| - image-text-to-text |
| tags: |
| - computer-use |
| - vla |
| - gui-agent |
| size_categories: |
| - 100K<n<1M |
| configs: |
| - config_name: default |
| data_files: |
| - split: train |
| path: data/train-* |
| dataset_info: |
| features: |
| - name: id |
| dtype: string |
| - name: source |
| dtype: string |
| - name: os_type |
| dtype: string |
| - name: episode_id |
| dtype: string |
| - name: step_idx |
| dtype: int32 |
| - name: image |
| dtype: image |
| - name: instruction |
| dtype: string |
| - name: action_type |
| dtype: int32 |
| - name: mouse_x |
| dtype: float32 |
| - name: mouse_y |
| dtype: float32 |
| - name: click_type |
| dtype: int32 |
| - name: keyboard_text |
| dtype: string |
| splits: |
| - name: train |
| num_bytes: 29315104297 |
| num_examples: 328834 |
| download_size: 29199843159 |
| dataset_size: 29315104297 |
| --- |
| |
| # TESS AgentNet Dataset |
|
|
| Computer use trajectories for training Vision-Language-Action models. |
|
|
| ## Dataset Description |
|
|
| - **Samples:** 328,834 |
| - **Source:** [AgentNet](https://huggingface.co/datasets/xlangai/AgentNet) |
| - **Format:** WebP images + action annotations |
|
|
| ## Features |
|
|
| - `image`: Screenshot (PIL Image) |
| - `instruction`: Task description |
| - `action_type`: 0=MOUSE, 1=KEYBOARD |
| - `mouse_x`, `mouse_y`: Normalized coordinates [0,1] |
| - `click_type`: 0-8 (NO_CLICK, LEFT_CLICK, etc.) |
| - `keyboard_text`: Text with special tokens |
| - `os_type`: ubuntu, windows_macos |
| - `episode_id`, `step_idx`: Episode structure |
| |
| ## Click Types |
| |
| | Index | Type | Description | |
| |-------|------|-------------| |
| | 0 | NO_CLICK | Mouse move only | |
| | 1 | LEFT_CLICK | Left mouse button | |
| | 2 | RIGHT_CLICK | Right mouse button | |
| | 3 | DOUBLE_CLICK | Double left click | |
| | 4 | MIDDLE_CLICK | Middle mouse button | |
| | 5 | SCROLL_UP | Scroll wheel up | |
| | 6 | SCROLL_DOWN | Scroll wheel down | |
| | 7 | DRAG_START | Begin drag operation | |
| | 8 | DRAG_END | End drag operation | |
|
|
| ## Action Types |
|
|
| | Index | Type | |
| |-------|------| |
| | 0 | MOUSE | |
| | 1 | KEYBOARD | |
|
|
|
|
| ## Usage |
|
|
| ```python |
| from datasets import load_dataset |
| |
| dataset = load_dataset("TESS-Computer/tess-agentnet-clean", streaming=True) |
| |
| Citation |
| |
| Based on AgentNet from OpenCUA project. |
| |
