lsalomme/so101-d2v1-case-open

SO-101 flip open a headphone case (D2, v1)

Teleoperated demonstrations on a LeRobot SO-101 arm: flip open a closed clamshell headphone case whose zipper is already open (the task does not include unzipping). The gripper stays closed and acts as a finger: it gets under the lid's edge and lifts it open around the hinge. No pick-and-place. Recorded with LeRobot (dataset format v3.0).

Part of a set recorded with the same arm, cameras and lighting: so101-d1v2-cube-pick-place (D1) and so101-d3v1-pen-cup (D3).

Summary

Task string	Flip open the unzipped headphone case
Episodes	50 (the case starts closed in every episode)
Frames	15,409 at 30 fps (7.9–17.1 s per episode, 9.6 s median)
Robot	SO-101 follower (so_follower), driven by an SO-101 leader arm
Action / state	6 joint positions: shoulder pan, shoulder lift, elbow flex, wrist flex, wrist roll, gripper
Cameras	front: fixed scene view, GC2083 sensor, 1280×720 · wrist: on the gripper, GC0308 sensor, 640×480 · both MJPG, 30 fps

Recording setup

• Lighting: two strong work spotlights (construction-site floodlights) lighting the workspace.
• Exposure: fixed and short to avoid motion blur — front 5 ms, wrist 10 ms.
• White balance: front locked at 4000 K; wrist on auto (no fixed value renders neutral colours on this sensor), so wrist colours can vary slightly between episodes.
• Variation: the case's position and orientation vary between episodes. The lid was closed by hand between episodes.

Known limitations

• Episodes 17 and 36 include a retry: the lid falls back closed after a first lift and is opened again, so they are longer (17.1 s and 15.8 s). Kept as recovery demonstrations.
• Episode 0: the gripper opens slightly (up to ~7 %) for about 3 s; it stays closed in all other episodes.
• 15 episodes start or end with the arm idle for a few frames (5–7 % of the episode).

How to use

from lerobot.datasets import LeRobotDataset

ds = LeRobotDataset("lsalomme/so101-d2v1-case-open")
frame = ds[0]
frame["observation.images.front"]  # (3, 720, 1280) float tensor
frame["observation.images.wrist"]  # (3, 480, 640)
frame["observation.state"]          # (6,) joint positions
frame["action"]                     # (6,) leader-arm joint targets
frame["task"]                       # "Flip open the unzipped headphone case"

Train an ACT policy with LeRobot:

lerobot-train \
  --dataset.repo_id=lsalomme/so101-d2v1-case-open \
  --policy.type=act \
  --output_dir=outputs/train/act_so101_d2v1 \
  --policy.push_to_hub=false

Tested with lerobot 0.6.1.

Intended use

Imitation learning on an articulated, non-prehensile task, and few-shot transfer tests (e.g. training on 10 or 25 of the 50 episodes) alongside the cube and pen datasets.

Dataset Structure

meta/info.json:

{
    "codebase_version": "v3.0",
    "fps": 30,
    "features": {
        "action": {
            "dtype": "float32",
            "names": [
                "shoulder_pan.pos",
                "shoulder_lift.pos",
                "elbow_flex.pos",
                "wrist_flex.pos",
                "wrist_roll.pos",
                "gripper.pos"
            ],
            "shape": [
                6
            ]
        },
        "observation.state": {
            "dtype": "float32",
            "names": [
                "shoulder_pan.pos",
                "shoulder_lift.pos",
                "elbow_flex.pos",
                "wrist_flex.pos",
                "wrist_roll.pos",
                "gripper.pos"
            ],
            "shape": [
                6
            ]
        },
        "observation.images.front": {
            "dtype": "video",
            "shape": [
                720,
                1280,
                3
            ],
            "names": [
                "height",
                "width",
                "channels"
            ],
            "info": {
                "is_depth_map": false,
                "video.height": 720,
                "video.width": 1280,
                "video.codec": "av1",
                "video.pix_fmt": "yuv420p",
                "video.fps": 30,
                "video.channels": 3,
                "has_audio": false,
                "video.g": 2,
                "video.crf": 30,
                "video.preset": 12,
                "video.fast_decode": 0,
                "video.video_backend": "pyav",
                "video.extra_options": {}
            }
        },
        "observation.images.wrist": {
            "dtype": "video",
            "shape": [
                480,
                640,
                3
            ],
            "names": [
                "height",
                "width",
                "channels"
            ],
            "info": {
                "is_depth_map": false,
                "video.height": 480,
                "video.width": 640,
                "video.codec": "av1",
                "video.pix_fmt": "yuv420p",
                "video.fps": 30,
                "video.channels": 3,
                "has_audio": false,
                "video.g": 2,
                "video.crf": 30,
                "video.preset": 12,
                "video.fast_decode": 0,
                "video.video_backend": "pyav",
                "video.extra_options": {}
            }
        },
        "timestamp": {
            "dtype": "float32",
            "shape": [
                1
            ],
            "names": null
        },
        "frame_index": {
            "dtype": "int64",
            "shape": [
                1
            ],
            "names": null
        },
        "episode_index": {
            "dtype": "int64",
            "shape": [
                1
            ],
            "names": null
        },
        "index": {
            "dtype": "int64",
            "shape": [
                1
            ],
            "names": null
        },
        "task_index": {
            "dtype": "int64",
            "shape": [
                1
            ],
            "names": null
        }
    },
    "total_episodes": 50,
    "total_frames": 15409,
    "total_tasks": 1,
    "chunks_size": 1000,
    "data_files_size_in_mb": 100,
    "video_files_size_in_mb": 200,
    "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",
    "robot_type": "so_follower",
    "splits": {
        "train": "0:50"
    }
}

Citation

@misc{salommez2026so101d2v1,
  title        = {SO-101 flip open a headphone case (D2, v1)},
  author       = {Salommez, Luc},
  year         = {2026},
  publisher    = {Hugging Face},
  howpublished = {\url{https://huggingface.co/datasets/lsalomme/so101-d2v1-case-open}}
}