Grigorij/Tello_multifruit_sum_two_state_vars

This dataset was created using LeRobot.

Dataset Description

Repaired copy of Grigorij/Tello_multifruit_sum (Tello drone, 149 episodes, 31624 frames, 30 fps). Video, actions, tasks, timestamps and all indices are byte-for-byte identical to the original; only observation.state changed.

Why: in the original, observation.state was declared with shape=[1], which LeRobot deliberately stores as a bare scalar Arrow column (see get_hf_features_from_features). Policies that load several observation timesteps then receive a 2-D tensor [B, T] with no feature axis, and mistake the time axis for the feature axis. With vla_jepa (num_video_frames=8) this crashes in the action head: RuntimeError: mat1 and mat2 shapes cannot be multiplied (64x8 and 1x1024).

Fix: observation.state is now a genuine 2-channel vector, shape=[2], stored as fixed_size_list<float>[2], so it always carries a real feature axis:

channel	name	units	definition
0	height	cm	the original altitude, copied verbatim and unchanged
1	height_velocity	cm/s	backward finite difference of channel 0 within each episode: v[0] = 0, v[t] = (h[t] - h[t-1]) * 30

Channel 0 is bitwise identical to the original observation.state for all 31624 frames, so the original signal is fully recoverable with state[..., 0].

Caveat on channel 1: the Tello reports altitude quantised to 10 cm, so the finite difference is 0 for 97.8% of frames and exactly +/-300 cm/s otherwise. It marks vertical transitions rather than providing a smooth velocity estimate.

Built with lerobot.datasets.dataset_tools (remove_feature, then add_features, then recompute_stats); meta/stats.json and the per-episode stats in meta/episodes/ were both regenerated for the new 2-channel shape.

• Homepage: [More Information Needed]
• Paper: [More Information Needed]
• License: apache-2.0

Dataset Structure

meta/info.json:

{
    "codebase_version": "v3.0",
    "fps": 30,
    "features": {
        "observation.images.camera_front": {
            "dtype": "video",
            "shape": [
                480,
                640,
                3
            ],
            "names": [
                "height",
                "width",
                "channels"
            ],
            "info": {
                "video.height": 480,
                "video.width": 640,
                "video.codec": "av1",
                "video.pix_fmt": "yuv420p",
                "video.is_depth_map": false,
                "video.fps": 30,
                "video.channels": 3,
                "has_audio": false
            }
        },
        "action": {
            "dtype": "float32",
            "shape": [
                4
            ],
            "names": [
                "forward",
                "right",
                "up",
                "yaw"
            ]
        },
        "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
        },
        "observation.state": {
            "dtype": "float32",
            "shape": [
                2
            ],
            "names": [
                "height",
                "height_velocity"
            ]
        }
    },
    "total_episodes": 149,
    "total_frames": 31624,
    "total_tasks": 3,
    "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": null,
    "splits": {
        "train": "0:149"
    }
}

Citation

BibTeX:

[More Information Needed]