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: CastError
Message: Couldn't cast
schema_version: int64
artifact_type: string
privacy: string
video_id: string
frames: struct<detector_version: string, frame_idx: int64, hands_present: int64, left_confidence: double, le (... 225 chars omitted)
child 0, detector_version: string
child 1, frame_idx: int64
child 2, hands_present: int64
child 3, left_confidence: double
child 4, left_landmarks: list<item: double>
child 0, item: double
child 5, left_track_id: int64
child 6, right_confidence: double
child 7, right_landmarks: list<item: double>
child 0, item: double
child 8, right_track_id: int64
child 9, segment_frame_idx: int64
child 10, segment_id: int64
child 11, timestamp_ms: int64
child 12, video_id: string
segments: list<item: struct<segment_id: int64, summary: struct<text: string, steps: list<item: string>>, activ (... 914 chars omitted)
child 0, item: struct<segment_id: int64, summary: struct<text: string, steps: list<item: string>>, activity_observa (... 902 chars omitted)
child 0, segment_id: int64
child 1, summary: struct<text: string, steps: list<item: string>>
child 0, text: string
child 1, steps: list<item: string>
child 0, item: string
child 2, activity_observations: list<item: struct<window_idx: int64, caption: string, segment_id: int64, start_ms: int64, end_ms: in (... 35 chars omitted)
child 0, item: struct<window_idx: int64, caption: string, segment_id: int64, start_ms: int64, end_ms: int64, end_cl
...
child 0, verb: string
child 1, object: string
child 2, target: string
child 3, contact_type: string
child 4, visible: bool
child 5, caption: string
child 4, tool_in_use: string
child 5, coordination: string
child 6, handover_event: bool
child 7, scene: struct<>
child 8, source_schema_ok: bool
child 9, source_window_indices: list<item: int64>
child 0, item: int64
child 10, source_actions: list<item: struct<window_idx: int64, action_idx: int64>>
child 0, item: struct<window_idx: int64, action_idx: int64>
child 0, window_idx: int64
child 1, action_idx: int64
child 11, event_idx: int64
child 12, segment_id: int64
child 13, start_ms: int64
child 14, end_ms: int64
child 15, end_clamped_to_segment: bool
child 4, merge_policy: struct<name: string, max_gap_ms: int64, never_crosses_manual_cuts: bool, does_not_move_or_snap_bound (... 12 chars omitted)
child 0, name: string
child 1, max_gap_ms: int64
child 2, never_crosses_manual_cuts: bool
child 3, does_not_move_or_snap_boundaries: bool
model_id: string
-- schema metadata --
pandas: '{"index_columns": [], "column_indexes": [], "columns": [{"name":' + 721
to
{'schema_version': Value('int64'), 'artifact_type': Value('string'), 'privacy': Value('string'), 'video_id': Value('string'), 'model_id': Value('string'), 'segments': List({'segment_id': Value('int64'), 'summary': {'text': Value('string'), 'steps': List(Value('string'))}, 'activity_observations': List({'window_idx': Value('int64'), 'caption': Value('string'), 'segment_id': Value('int64'), 'start_ms': Value('int64'), 'end_ms': Value('int64'), 'end_clamped_to_segment': Value('bool')}), 'events': List({'task_step': Value('string'), 'action_caption': Value('string'), 'left_hand': {'verb': Value('string'), 'object': Value('string'), 'target': Value('string'), 'contact_type': Value('string'), 'visible': Value('bool'), 'caption': Value('string')}, 'right_hand': {'verb': Value('string'), 'object': Value('string'), 'target': Value('string'), 'contact_type': Value('string'), 'visible': Value('bool'), 'caption': Value('string')}, 'tool_in_use': Value('string'), 'coordination': Value('string'), 'handover_event': Value('bool'), 'scene': {}, 'source_schema_ok': Value('bool'), 'source_window_indices': List(Value('int64')), 'source_actions': List({'window_idx': Value('int64'), 'action_idx': Value('int64')}), 'event_idx': Value('int64'), 'segment_id': Value('int64'), 'start_ms': Value('int64'), 'end_ms': Value('int64'), 'end_clamped_to_segment': Value('bool')}), 'merge_policy': {'name': Value('string'), 'max_gap_ms': Value('int64'), 'never_crosses_manual_cuts': Value('bool'), 'does_not_move_or_snap_boundaries': Value('bool')}})}
because column names don't match
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 290, in _generate_tables
pa_table = paj.read_json(
io.BytesIO(batch), read_options=paj.ReadOptions(block_size=block_size)
)
File "pyarrow/_json.pyx", line 342, in pyarrow._json.read_json
File "pyarrow/error.pxi", line 155, in pyarrow.lib.pyarrow_internal_check_status
File "pyarrow/error.pxi", line 92, in pyarrow.lib.check_status
raise convert_status(status)
pyarrow.lib.ArrowInvalid: JSON parse error: Invalid value. in row 0
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/src/services/worker/src/worker/utils.py", line 147, 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 127, 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 483, 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 2840, in __iter__
for key, example in ex_iterable:
^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2373, in __iter__
for key, pa_table in self._iter_arrow():
~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2398, in _iter_arrow
for key, pa_table in self.ex_iterable._iter_arrow():
~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 536, in _iter_arrow
for key, pa_table in iterator:
^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 419, in _iter_arrow
for key, pa_table in self.generate_tables_fn(**gen_kwags):
~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 339, in _generate_tables
yield Key(shard_idx, 0), self._cast_table(pa_table)
~~~~~~~~~~~~~~~~^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
pa_table = table_cast(pa_table, self.info.features.arrow_schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2378, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2306, in cast_table_to_schema
raise CastError(
...<3 lines>...
)
datasets.table.CastError: Couldn't cast
schema_version: int64
artifact_type: string
privacy: string
video_id: string
frames: struct<detector_version: string, frame_idx: int64, hands_present: int64, left_confidence: double, le (... 225 chars omitted)
child 0, detector_version: string
child 1, frame_idx: int64
child 2, hands_present: int64
child 3, left_confidence: double
child 4, left_landmarks: list<item: double>
child 0, item: double
child 5, left_track_id: int64
child 6, right_confidence: double
child 7, right_landmarks: list<item: double>
child 0, item: double
child 8, right_track_id: int64
child 9, segment_frame_idx: int64
child 10, segment_id: int64
child 11, timestamp_ms: int64
child 12, video_id: string
segments: list<item: struct<segment_id: int64, summary: struct<text: string, steps: list<item: string>>, activ (... 914 chars omitted)
child 0, item: struct<segment_id: int64, summary: struct<text: string, steps: list<item: string>>, activity_observa (... 902 chars omitted)
child 0, segment_id: int64
child 1, summary: struct<text: string, steps: list<item: string>>
child 0, text: string
child 1, steps: list<item: string>
child 0, item: string
child 2, activity_observations: list<item: struct<window_idx: int64, caption: string, segment_id: int64, start_ms: int64, end_ms: in (... 35 chars omitted)
child 0, item: struct<window_idx: int64, caption: string, segment_id: int64, start_ms: int64, end_ms: int64, end_cl
...
child 0, verb: string
child 1, object: string
child 2, target: string
child 3, contact_type: string
child 4, visible: bool
child 5, caption: string
child 4, tool_in_use: string
child 5, coordination: string
child 6, handover_event: bool
child 7, scene: struct<>
child 8, source_schema_ok: bool
child 9, source_window_indices: list<item: int64>
child 0, item: int64
child 10, source_actions: list<item: struct<window_idx: int64, action_idx: int64>>
child 0, item: struct<window_idx: int64, action_idx: int64>
child 0, window_idx: int64
child 1, action_idx: int64
child 11, event_idx: int64
child 12, segment_id: int64
child 13, start_ms: int64
child 14, end_ms: int64
child 15, end_clamped_to_segment: bool
child 4, merge_policy: struct<name: string, max_gap_ms: int64, never_crosses_manual_cuts: bool, does_not_move_or_snap_bound (... 12 chars omitted)
child 0, name: string
child 1, max_gap_ms: int64
child 2, never_crosses_manual_cuts: bool
child 3, does_not_move_or_snap_boundaries: bool
model_id: string
-- schema metadata --
pandas: '{"index_columns": [], "column_indexes": [], "columns": [{"name":' + 721
to
{'schema_version': Value('int64'), 'artifact_type': Value('string'), 'privacy': Value('string'), 'video_id': Value('string'), 'model_id': Value('string'), 'segments': List({'segment_id': Value('int64'), 'summary': {'text': Value('string'), 'steps': List(Value('string'))}, 'activity_observations': List({'window_idx': Value('int64'), 'caption': Value('string'), 'segment_id': Value('int64'), 'start_ms': Value('int64'), 'end_ms': Value('int64'), 'end_clamped_to_segment': Value('bool')}), 'events': List({'task_step': Value('string'), 'action_caption': Value('string'), 'left_hand': {'verb': Value('string'), 'object': Value('string'), 'target': Value('string'), 'contact_type': Value('string'), 'visible': Value('bool'), 'caption': Value('string')}, 'right_hand': {'verb': Value('string'), 'object': Value('string'), 'target': Value('string'), 'contact_type': Value('string'), 'visible': Value('bool'), 'caption': Value('string')}, 'tool_in_use': Value('string'), 'coordination': Value('string'), 'handover_event': Value('bool'), 'scene': {}, 'source_schema_ok': Value('bool'), 'source_window_indices': List(Value('int64')), 'source_actions': List({'window_idx': Value('int64'), 'action_idx': Value('int64')}), 'event_idx': Value('int64'), 'segment_id': Value('int64'), 'start_ms': Value('int64'), 'end_ms': Value('int64'), 'end_clamped_to_segment': Value('bool')}), 'merge_policy': {'name': Value('string'), 'max_gap_ms': Value('int64'), 'never_crosses_manual_cuts': Value('bool'), 'does_not_move_or_snap_boundaries': Value('bool')}})}
because column names don't matchNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
EgoAnnotate v1
Built by Taher Panbiharwala, with Zainab Barwaniwala.
This public release contains 13 manually reviewed face-free egocentric videos with synchronized MediaPipe hand landmarks and dense activity captions. Each clip includes its clean video, a hand-and-caption overlay video, hand data as JSON and Parquet, and captions as JSON.
The full annotation pipeline — curation, hand tracking, dense captioning, and event compilation — is open-sourced at github.com/TaherPanbiharwala/EgoAnnotate.
Curation and continuity
The clean videos are normalized to 29.97 fps with audio and container metadata removed. Every per-video manifest declares its annotation segments and any privacy-curation discontinuities. At a discontinuity, MediaPipe track state and caption windows restart; consumers must not treat annotations across it as one continuous track. This v1 release has no frame-removal discontinuities: each clip is a single, manually reviewed face-free segment.
Files
- videos/: clean, normalized videos.
- annotated-videos/: final videos with hand landmarks and captions burned in.
- annotations/video-id/hands.json: one hand annotation record per output frame.
- annotations/video-id/hands.parquet: the same records in columnar form.
- annotations/video-id/captions.json: summaries, activity captions, atomic actions, and structured left/right-hand states.
- manifests/: public media hashes, curation metadata, and annotation counts.
Limitations
This is a small, owner-curated release. The captions are model-generated and should be treated as annotations rather than ground truth. The per-video manifest is the source of truth for curation boundaries and annotation continuity.
Citation
If you use this dataset, please cite:
@dataset{panbiharwala2026egoannotate,
author = {Panbiharwala, Taher and Barwaniwala, Zainab},
title = {EgoAnnotate: A Dense Multi-Level Annotation Dataset for Egocentric Video},
year = {2026},
publisher = {Hugging Face},
url = {https://huggingface.co/datasets/TaherPanbiharwala/EgoAnnotate}
}
- Downloads last month
- 141