vvm
Collection
6 items • Updated
Dataset Viewer
The dataset viewer is not available for this split.
Cannot load the dataset split (in streaming mode) to extract the first rows.
Error code: StreamingRowsError
Exception: ValueError
Message: Invalid string class label data
Traceback: Traceback (most recent call last):
File "/src/services/worker/src/worker/utils.py", line 149, in get_rows_or_raise
return get_rows(
dataset=dataset,
...<4 lines>...
column_names=column_names,
)
File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
return func(*args, **kwargs)
File "/src/services/worker/src/worker/utils.py", line 129, in get_rows
rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
File "/src/services/worker/src/worker/utils.py", line 544, in safe_iter
yield from ds.decode(False) if ds.features else ds
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 3052, in __iter__
for key, example in ex_iterable:
^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2572, in __iter__
example = _apply_feature_types_on_example(
example, self.features, token_per_repo_id=self.token_per_repo_id
)
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2489, in _apply_feature_types_on_example
encoded_example = features.encode_example(example)
File "/usr/local/lib/python3.14/site-packages/datasets/features/features.py", line 2244, in encode_example
return encode_nested_example(self, example)
File "/usr/local/lib/python3.14/site-packages/datasets/features/features.py", line 1518, in encode_nested_example
{k: encode_nested_example(schema[k], obj.get(k), level=level + 1) for k in schema}
~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/features/features.py", line 1541, in encode_nested_example
return schema.encode_example(obj) if obj is not None else None
~~~~~~~~~~~~~~~~~~~~~^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/features/features.py", line 1210, in encode_example
example_data = self.str2int(example_data)
File "/usr/local/lib/python3.14/site-packages/datasets/features/features.py", line 1147, in str2int
output = [self._strval2int(value) for value in values]
~~~~~~~~~~~~~~~~^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/features/features.py", line 1168, in _strval2int
raise ValueError(f"Invalid string class label {value}")
ValueError: Invalid string class label dataNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
DROID videos (H.264)
The camera videos of DROID, repackaged as plain H.264 mp4 files for
video-model training. The source is the LeRobot port
cadene/droid at revision
c8fbc029c9786fd4377ea20d9535e86d88199c0f. Only the videos are kept; the per-frame robot
state and action parquet files are not included.
- 92,233 episodes x 3 camera views = 276,699 videos
- 320x180, 15 fps; resolution, frame rate and frame count unchanged from the source
- The source AV1 videos were re-encoded to H.264 (libx264, CRF 18, yuv420p, no audio) so that decoders without AV1 support (e.g. decord) can read them. Every output video's frame count was checked against the source episode metadata: 0 failures, 0 mismatches.
Layout
data/data.json [{id, video_path, split, category}], one entry per video
data/<view>-NN.zip 10,000 episodes per archive (stored, not compressed)
members: <view>/episode_NNNNNN.mp4
manifests/<view>/<setting>/ clip manifests (manifest.json + meta.yaml)
<view> is one of exterior_image_1_left, exterior_image_2_left, wrist_image_left.
In data.json, split is the view and category is the episode's task text (empty when the
source has none). The same episode index refers to the same episode in all three views.
Download and extract:
hf download stzhao/droid_video --repo-type dataset --local-dir droid_video
cd droid_video/data && for z in *.zip; do unzip -q "$z"; done
After extraction, each data/<view>/ directory is a video root, and the manifests' paths are
relative to it.
Clip manifests
Each manifest lists (video, start frame) records: every video is cut from frame 0 into
non-overlapping windows of source_frame_span = 1 + (rgb_seq_len - 1) * frame_interval frames,
and the incomplete tail is discarded. The three views have identical statistics because their
videos have the same lengths.
| Setting | Frames / interval | Span | Usable videos (of 92,233) | Clips per view |
|---|---|---|---|---|
cond1gen15_interval1 |
16 / 1 | 16 | 91,854 | 1,647,121 |
cond1gen15_interval3 |
16 / 3 | 46 | 90,992 | 542,860 |
cond1gen16_interval3 |
17 / 3 | 49 | 90,912 | 506,848 |
License and citation
DROID is released under Apache 2.0. If you use these videos, please cite the original dataset:
@article{khazatsky2024droid,
title={Droid: A large-scale in-the-wild robot manipulation dataset},
author={Khazatsky, Alexander and Pertsch, Karl and Nair, Suraj and Balakrishna, Ashwin and Dasari, Sudeep and Karamcheti, Siddharth and Nasiriany, Soroush and Srirama, Mohan Kumar and Chen, Lawrence Yunliang and Ellis, Kirsty and others},
journal={arXiv preprint arXiv:2403.12945},
year={2024}
}
- Downloads last month
- 101