| configs: | |
| - config_name: default | |
| data_files: | |
| - split: train | |
| path: data/**/*.parquet | |
| # Trimmed Towel Folding Dataset (Multi-Stream) | |
| This is a trimmed version of arsenxeno/augmented-setup where each episode is cut at the first fold detection. | |
| All video streams are trimmed to the same frame based on detection from a single stream. | |
| Fold detection performed using: observation.images.front | |
| ## Dataset Info | |
| - Total episodes: 1 | |
| - Total frames: 337 | |
| - Total videos: 3 | |
| - Structure: chunked (chunk_size=50) | |
| - Episodes included: [0] | |
| ## Usage with LeRobot | |
| ```python | |
| from lerobot.common.datasets.lerobot_dataset import LeRobotDataset | |
| dataset = LeRobotDataset( | |
| repo_id="xenorobotics/new-28", | |
| version="v2.0" | |
| ) | |
| ``` | |
| ## Training | |
| ```bash | |
| lerobot train \ | |
| dataset_repo_id=xenorobotics/new-28 \ | |
| policy=diffusion \ | |
| env=so100_real \ | |
| hydra.run.dir=outputs/train | |
| ``` | |
| ## Original Dataset | |
| Source: https://huggingface.co/datasets/arsenxeno/augmented-setup | |