Delcastillo8/libero90_frying_pan_phone_50

This dataset was created using LeRobot.

Dataset Description

Robot demonstrations of LIBERO-90 task 21, "turn on the stove and put the frying pan on it", made from phone videos of a person moving a real frying pan in a real kitchen.

How each demo was made (code and write-up):

1. A printed marker inside a real pan and a printed board on the stove were filmed with an iPad (18 takes, 17 of which gave demos). The pan's 3-D path was tracked in every frame.
2. Each path was fitted onto LIBERO's kitchen (rotated, scaled, shifted onto the simulated pan and burner; played 1.5x faster; raised locally to clear the moka pot).
3. A scripted Franka Panda in LIBERO turned the stove knob, grasped the pan's handle and carried the pan along the path, with physics; only episodes that passed LIBERO's own success check and quality gates (path followed within 2 cm RMS, moka pot undisturbed) were kept.
4. Recorded in the conventions of HuggingFaceVLA/libero (the data smolvla_libero was trained on): observation.images.image (front) and observation.images.image2 (wrist), 256x256, rotated 180 degrees as LeRobot's LIBERO processor does; observation.state 8-D (end-effector position, axis-angle, gripper joints); action 7-D; fps label 10, one frame per 20 Hz control step. Frames are stored as AV1 video (HuggingFaceVLA/libero stores images in the parquet files; LeRobot reads both).

50 episodes, 13811 frames. The knob turn and the grasp are scripted; the carry (where the pan goes, how high, how it turns, how fast) comes from the phone videos. None of the episodes starts from LIBERO's 50 fixed evaluation layouts (nearest: 0.3-2.5 cm).

The 50 demos of run P (Delcastillo8/smolvla_libero_frying_pan_P), taken round-robin over the 17 takes. All 153 demos (run Pall) are in Delcastillo8/libero90_frying_pan_phone.

• Homepage: https://github.com/pd05849/HumanoidAI_challenge
• License: apache-2.0

Dataset Structure

meta/info.json:

{
    "codebase_version": "v3.0",
    "fps": 10,
    "features": {
        "observation.images.image": {
            "dtype": "video",
            "shape": [
                256,
                256,
                3
            ],
            "names": [
                "height",
                "width",
                "channels"
            ],
            "info": {
                "video.height": 256,
                "video.width": 256,
                "video.codec": "av1",
                "video.pix_fmt": "yuv420p",
                "video.fps": 10,
                "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": {},
                "is_depth_map": false
            }
        },
        "observation.images.image2": {
            "dtype": "video",
            "shape": [
                256,
                256,
                3
            ],
            "names": [
                "height",
                "width",
                "channels"
            ],
            "info": {
                "video.height": 256,
                "video.width": 256,
                "video.codec": "av1",
                "video.pix_fmt": "yuv420p",
                "video.fps": 10,
                "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": {},
                "is_depth_map": false
            }
        },
        "observation.state": {
            "dtype": "float32",
            "shape": [
                8
            ],
            "names": null
        },
        "action": {
            "dtype": "float32",
            "shape": [
                7
            ],
            "names": null
        },
        "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": 13811,
    "total_tasks": 1,
    "chunks_size": 1000,
    "data_files_size_in_mb": 90,
    "video_files_size_in_mb": 90,
    "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": "panda",
    "splits": {
        "train": "0:50"
    }
}