| --- |
| pretty_name: Sim1 Dataset (LeRobot Format) |
| license: apache-2.0 |
| task_categories: |
| - robotics |
| language: |
| - en |
| tags: |
| - lerobot |
| - embodied-ai |
| - robotic-manipulation |
| - imitation-learning |
| size_categories: |
| - 10K<n<100K |
| configs: |
| - config_name: default |
| data_files: |
| - split: train |
| path: |
| - "real_three_fold/data/**/*.parquet" |
| - "scale_fold_mat/data/**/*.parquet" |
| - "scale_shorts/data/**/*.parquet" |
| - "scale_fold_shirt_in_domain/data/**/*.parquet" |
| - "scale_fold_shirt_randomtexture/data/**/*.parquet" |
| - "scale_fold_shirt_randpos/data/**/*.parquet" |
| --- |
| |
| # Sim1_Dataset |
| |
| LeRobot-format manipulation dataset for cloth folding / deformable object tasks. |
| |
| ## Format |
| |
| This repository follows a LeRobot-style layout per subset: |
| |
| - `data/chunk-xxx/episode_XXXXXX.parquet`: frame-level trajectories |
| - `videos/chunk-xxx/images.rgb.{head,hand_left,hand_right}/episode_XXXXXX.mp4`: multi-view videos |
| - `meta/episodes.jsonl`, `meta/episodes_stats.jsonl`, `meta/tasks.jsonl`, `meta/info.json`: episode/task metadata |
|
|
| The **Dataset Viewer** is configured from the parquet files in each subset via the `configs.data_files` section above. |
|
|
| ## Subsets (configs) |
|
|
| - `real_three_fold` |
| - `scale_fold_mat` |
| - `scale_shorts` |
| - `scale_fold_shirt_in_domain` |
| - `scale_fold_shirt_randomtexture` |
| - `scale_fold_shirt_randpos` |
|
|
| ## Notes |
|
|
| - Primary supervised signal is from parquet trajectories. |
| - MP4 files are provided for visualization and replay. |
| - If access is gated, file listing is visible but content requires accepted access conditions. |
|
|
| ## 📚 Citation |
|
|
| If you find this work useful, please cite: |
|
|
| **SIM1** (arXiv:2604.08544) |
|
|
| https://arxiv.org/abs/2604.08544 |
|
|
