chaoqi-liu/bridge_simpler_mt4

---
license: apache-2.0
configs:
- config_name: default
data_files: data/*/*.parquet
tags:
- lerobot
- LeRobot
- robotics
- bridge
- simpler
task_categories:
- robotics
pretty_name: bridge_simpler_mt4
---

# bridge_simpler_mt4

This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).

## Dataset Description

This dataset is a targeted Bridge subset for the four SIMPLER WidowX tasks:

- `widowx_spoon_on_towel`
- `widowx_carrot_on_plate`
- `widowx_stack_cube`
- `widowx_put_eggplant_in_basket`

The data preserve the source Bridge camera, state, and action schema while narrowing the
task distribution toward these four targets.

## Dataset Structure

[meta/info.json](https://huggingface.co/datasets/chaoqi-liu/bridge_simpler_mt4/blob/main/meta/info.json):

```json
{
"codebase_version": "v3.0",
"robot_type": "widowx",
"total_episodes": 2237,
"total_frames": 83482,
"total_tasks": 1217,
"fps": 5,
"splits": {
"train": "0:2237"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4"
}
```

- [meta/stats.json](https://huggingface.co/datasets/chaoqi-liu/bridge_simpler_mt4/blob/main/meta/stats.json)
- [meta/tasks.parquet](https://huggingface.co/datasets/chaoqi-liu/bridge_simpler_mt4/blob/main/meta/tasks.parquet)
- [meta/episodes/chunk-000/file-000.parquet](https://huggingface.co/datasets/chaoqi-liu/bridge_simpler_mt4/blob/main/meta/episodes/chunk-000/file-000.parquet)

## Source

- Source repo: `jesbu1/bridge_v2_lerobot`
- Output repo: `chaoqi-liu/bridge_simpler_mt4`

## Selection Recipe

The subset is built online from source metadata with this frozen recipe:

- direct `spoon_on_towel`
- direct `carrot_on_plate`
- direct loose `stack_cube`
- compositional `eggplant_any`
- compositional `basket_any`
- loose `stack_any_blockcube`

This recipe is intentionally asymmetric:

- task 4 has no exact public Bridge match, so it is approximated with `eggplant_any ∪ basket_any`
- task 3 is widened with loose stack/block-cube tasks to raise subset size into a practical range

## Final Size

- episodes: `2237`
- frames: `83482`

## Rule Counts Before Union

- `spoon_direct`: `167`
- `carrot_direct`: `332`
- `stack_direct_loose`: `33`
- `eggplant_any`: `418`
- `basket_any`: `544`
- `stack_any_blockcube`: `772`

## Notes

- The dataset name is stable; exact counts are recorded here instead of being encoded in the repo name.
- This subset preserves the source camera/state/action schema.