The dataset viewer is not available for this split.
Error code: StreamingRowsError
Exception: CastError
Message: Couldn't cast
sample_id: string
language: string
asr: string
words: list<item: struct<word: string, start: double, end: double>>
child 0, item: struct<word: string, start: double, end: double>
child 0, word: string
child 1, start: double
child 2, end: double
max_gap_ms: struct<gyr: double, rot: double, acc: double>
child 0, gyr: double
child 1, rot: double
child 2, acc: double
clock: string
device: struct<manufacturer: string, model: string, platform: string, hardware_gyro: bool>
child 0, manufacturer: string
child 1, model: string
child 2, platform: string
child 3, hardware_gyro: bool
mount: struct<type: string, pitch_down_deg: int64>
child 0, type: string
child 1, pitch_down_deg: int64
samples: list<item: struct<s: string, t: int64, v: list<item: double>>>
child 0, item: struct<s: string, t: int64, v: list<item: double>>
child 0, s: string
child 1, t: int64
child 2, v: list<item: double>
child 0, item: double
counts: struct<acc: int64, gyr: int64, rot: int64>
child 0, acc: int64
child 1, gyr: int64
child 2, rot: int64
units: struct<acc: string, gyr: string, rot: string>
child 0, acc: string
child 1, gyr: string
child 2, rot: string
measured_rate_hz: struct<acc: double, gyr: double, rot: double>
child 0, acc: double
child 1, gyr: double
child 2, rot: double
camera: struct<ultrawide: bool, lens: struct<facing: string, focalLengthMm: double, horizontalFovDeg: double (... 161 chars omitted)
child 0, ultrawide: bool
child 1, lens: struct<facing: string, focalLengthMm: double, horizontalFovDeg: double, verticalFovDeg: double, diag (... 130 chars omitted)
child 0, facing: string
child 1, focalLengthMm: double
child 2, horizontalFovDeg: double
child 3, verticalFovDeg: double
child 4, diagonalFovDeg: double
child 5, isUltrawide: bool
child 6, intrinsicCalibration: list<item: double>
child 0, item: double
child 7, distortion: null
child 8, activeArray: list<item: int64>
child 0, item: int64
video_frame_timestamps_ns: list<item: int64>
child 0, item: int64
duration_s: double
schema: string
to
{'schema': Value('string'), 'sample_id': Value('string'), 'clock': Value('string'), 'units': {'acc': Value('string'), 'gyr': Value('string'), 'rot': Value('string')}, 'device': {'manufacturer': Value('string'), 'model': Value('string'), 'platform': Value('string'), 'hardware_gyro': Value('bool')}, 'camera': {'ultrawide': Value('bool'), 'lens': {'facing': Value('string'), 'focalLengthMm': Value('float64'), 'horizontalFovDeg': Value('float64'), 'verticalFovDeg': Value('float64'), 'diagonalFovDeg': Value('float64'), 'isUltrawide': Value('bool'), 'intrinsicCalibration': List(Value('float64')), 'distortion': Value('null'), 'activeArray': List(Value('int64'))}}, 'mount': {'type': Value('string'), 'pitch_down_deg': Value('int64')}, 'counts': {'acc': Value('int64'), 'gyr': Value('int64'), 'rot': Value('int64')}, 'measured_rate_hz': {'acc': Value('float64'), 'gyr': Value('float64'), 'rot': Value('float64')}, 'max_gap_ms': {'gyr': Value('float64'), 'rot': Value('float64'), 'acc': Value('float64')}, 'duration_s': Value('float64'), 'video_frame_timestamps_ns': List(Value('int64')), 'samples': List({'s': Value('string'), 't': Value('int64'), 'v': List(Value('float64'))})}
because column names don't match
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 2559, in __iter__
for key, pa_table in self._iter_arrow():
~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2584, 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 558, in _iter_arrow
for key, pa_table in iterator:
^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 440, 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 361, in _generate_tables
self._cast_table(pa_table, json_field_paths=json_field_paths),
~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
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 2381, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2309, in cast_table_to_schema
raise CastError(
...<3 lines>...
)
datasets.table.CastError: Couldn't cast
sample_id: string
language: string
asr: string
words: list<item: struct<word: string, start: double, end: double>>
child 0, item: struct<word: string, start: double, end: double>
child 0, word: string
child 1, start: double
child 2, end: double
max_gap_ms: struct<gyr: double, rot: double, acc: double>
child 0, gyr: double
child 1, rot: double
child 2, acc: double
clock: string
device: struct<manufacturer: string, model: string, platform: string, hardware_gyro: bool>
child 0, manufacturer: string
child 1, model: string
child 2, platform: string
child 3, hardware_gyro: bool
mount: struct<type: string, pitch_down_deg: int64>
child 0, type: string
child 1, pitch_down_deg: int64
samples: list<item: struct<s: string, t: int64, v: list<item: double>>>
child 0, item: struct<s: string, t: int64, v: list<item: double>>
child 0, s: string
child 1, t: int64
child 2, v: list<item: double>
child 0, item: double
counts: struct<acc: int64, gyr: int64, rot: int64>
child 0, acc: int64
child 1, gyr: int64
child 2, rot: int64
units: struct<acc: string, gyr: string, rot: string>
child 0, acc: string
child 1, gyr: string
child 2, rot: string
measured_rate_hz: struct<acc: double, gyr: double, rot: double>
child 0, acc: double
child 1, gyr: double
child 2, rot: double
camera: struct<ultrawide: bool, lens: struct<facing: string, focalLengthMm: double, horizontalFovDeg: double (... 161 chars omitted)
child 0, ultrawide: bool
child 1, lens: struct<facing: string, focalLengthMm: double, horizontalFovDeg: double, verticalFovDeg: double, diag (... 130 chars omitted)
child 0, facing: string
child 1, focalLengthMm: double
child 2, horizontalFovDeg: double
child 3, verticalFovDeg: double
child 4, diagonalFovDeg: double
child 5, isUltrawide: bool
child 6, intrinsicCalibration: list<item: double>
child 0, item: double
child 7, distortion: null
child 8, activeArray: list<item: int64>
child 0, item: int64
video_frame_timestamps_ns: list<item: int64>
child 0, item: int64
duration_s: double
schema: string
to
{'schema': Value('string'), 'sample_id': Value('string'), 'clock': Value('string'), 'units': {'acc': Value('string'), 'gyr': Value('string'), 'rot': Value('string')}, 'device': {'manufacturer': Value('string'), 'model': Value('string'), 'platform': Value('string'), 'hardware_gyro': Value('bool')}, 'camera': {'ultrawide': Value('bool'), 'lens': {'facing': Value('string'), 'focalLengthMm': Value('float64'), 'horizontalFovDeg': Value('float64'), 'verticalFovDeg': Value('float64'), 'diagonalFovDeg': Value('float64'), 'isUltrawide': Value('bool'), 'intrinsicCalibration': List(Value('float64')), 'distortion': Value('null'), 'activeArray': List(Value('int64'))}}, 'mount': {'type': Value('string'), 'pitch_down_deg': Value('int64')}, 'counts': {'acc': Value('int64'), 'gyr': Value('int64'), 'rot': Value('int64')}, 'measured_rate_hz': {'acc': Value('float64'), 'gyr': Value('float64'), 'rot': Value('float64')}, 'max_gap_ms': {'gyr': Value('float64'), 'rot': Value('float64'), 'acc': Value('float64')}, 'duration_s': Value('float64'), 'video_frame_timestamps_ns': List(Value('int64')), 'samples': List({'s': Value('string'), 't': Value('int64'), 'v': List(Value('float64'))})}
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.
Egocentric Home Activities with Narration and IMU (Sample)
A free sample of first-person video recorded at home, captured with a phone worn on a head strap. Every clip includes continuous spoken English narration, synchronised IMU data (accelerometer, gyroscope, attitude), camera lens parameters and word-level transcripts.
It is a small, hand-checked slice of a larger collection that Train Them AI records and licenses for training robotics, embodied-AI and video-language models.
| Clips | 48 |
| Total duration | 13.03 h |
| Clip length | 7 to 20 min, continuous, no cuts |
| Setting | private homes (West Africa) |
| Contributors | 11 adults |
| Video | 1920×1080, ~30 fps (41 H.264, 7 HEVC), original camera encode, ultrawide lens, head-mounted, about 45° down |
| Audio | narrator's own voice, English, 80 words/min median |
| Sensors | accelerometer, gyroscope and rotation vector, 100 Hz (iPhone) to 470 Hz (Android), hardware-timestamped against every video frame |
Activities
| Category | Clips | Hours | Share |
|---|---|---|---|
| arts | 17 | 4.41 | 34% |
| garment making | 9 | 2.56 | 20% |
| home repair | 5 | 1.35 | 10% |
| assembly | 5 | 1.20 | 9% |
| knitting & crochet | 3 | 0.99 | 8% |
| food preparation | 3 | 0.74 | 6% |
| paper & bookbinding | 2 | 0.66 | 5% |
| gardening | 2 | 0.46 | 4% |
| jewellery making | 1 | 0.33 | 3% |
| woodworking | 1 | 0.33 | 3% |
All clips
| Clip | Narrator | Category | Activity | Length |
|---|---|---|---|---|
home_001 |
narrator_A | arts | watercolour painting: a street scene with buses | 19.6 min |
home_003 |
narrator_A | arts | painting: a standing fan and wall socket | 13.5 min |
home_004 |
narrator_B | arts | drawing and colouring: a teddy bear, pencil and pen | 19.5 min |
home_005 |
narrator_C | arts | pencil drawing: a full human figure | 18.6 min |
home_006 |
narrator_D | garment making | machine-sewing sleeves onto a shirt and hand-sewing a button | 19.8 min |
home_007 |
narrator_E | food preparation | hand-sorting stored grain to remove weevils | 13.1 min |
home_008 |
narrator_E | food preparation | sieving stored grain to remove weevils | 13.1 min |
home_012 |
narrator_A | garment making | hand-sewing three buttons onto a t-shirt | 17.3 min |
home_013 |
narrator_A | garment making | hand-sewing buttons onto the front of a t-shirt | 15.5 min |
home_014 |
narrator_B | arts | pencil drawing and colouring: an ostrich | 19.6 min |
home_015 |
narrator_C | arts | fashion sketch: a top and trousers, pencil and ruler | 14.8 min |
home_016 |
narrator_C | arts | pencil drawing: a ladybug | 8.9 min |
home_018 |
narrator_C | arts | pencil drawing with a ruler: a handbag | 8.4 min |
home_019 |
narrator_D | garment making | sewing a striped pouch bag on a manual sewing machine | 19.8 min |
home_020 |
narrator_E | food preparation | washing, peeling and hand-grinding a spice blend | 18.2 min |
home_021 |
narrator_F | woodworking | sawing planks to length with a hand saw | 19.7 min |
home_022 |
narrator_G | arts | pen drawing and shading: a geometric polar bear | 18.7 min |
home_024 |
narrator_H | arts | watercolour painting: a rock study | 16.4 min |
home_025 |
narrator_H | arts | watercolour painting: leaves and ground | 19.0 min |
home_029 |
narrator_G | arts | pen drawing and colouring: a clothed figure | 14.6 min |
home_031 |
narrator_G | arts | ink drawing and colouring: a portrait figure | 14.3 min |
home_032 |
narrator_G | arts | ink drawing and colouring: a woman in a hat and red dress | 14.7 min |
home_033 |
narrator_G | arts | ink drawing and colouring: a cartoon character | 14.8 min |
home_035 |
narrator_I | arts | watercolour painting: a sky and sun | 14.8 min |
home_036 |
narrator_I | arts | watercolour painting: a landscape with sun and water | 14.6 min |
home_037 |
narrator_J | assembly | assembling adjustable dumbbells: weight plates, bars and collars | 14.0 min |
home_038 |
narrator_B | assembly | assembling a plastic-and-rod shoe rack | 18.7 min |
home_039 |
narrator_E | assembly | assembling a metal shoe rack, taping and tightening joints | 20.0 min |
home_040 |
narrator_F | assembly | finishing a shelf: fitting metal bases into the legs | 11.6 min |
home_041 |
narrator_C | assembly | assembling a multi-tier plastic storage cabinet | 7.4 min |
home_042 |
narrator_B | home repair | taking apart and repairing a hair clipper, rewiring the lead | 19.8 min |
home_043 |
narrator_J | home repair | reassembling and reconnecting three standing fans and a portable fan | 11.7 min |
home_044 |
narrator_J | home repair | reassembling and plugging in two electric fans | 10.3 min |
home_045 |
narrator_F | home repair | stripping and cleaning a petrol generator: fuel tank and parts | 19.6 min |
home_046 |
narrator_E | home repair | assembling a ring-light stand and phone holder | 19.8 min |
home_047 |
narrator_A | garment making | hand-sewing three buttons onto a shirt | 16.9 min |
home_048 |
narrator_C | garment making | hand-stitching a torn seam on a shirt, outdoors at home | 12.8 min |
home_049 |
narrator_K | garment making | hand-stitching a fabric appliqué onto a sleeve | 19.6 min |
home_050 |
narrator_D | garment making | hand-sewing a blind hem on pink fabric | 18.7 min |
home_051 |
narrator_D | garment making | hand-sewing an invisible hem on white fabric | 12.8 min |
home_052 |
narrator_B | knitting & crochet | hand-knitting rows of a shawl on two needles | 19.8 min |
home_053 |
narrator_A | knitting & crochet | crocheting a long chain in purple yarn | 19.6 min |
home_054 |
narrator_A | knitting & crochet | crocheting a chain-stitch bracelet | 19.8 min |
home_055 |
narrator_E | jewellery making | stringing and knotting a beaded bracelet | 19.8 min |
home_056 |
narrator_F | paper & bookbinding | trimming, folding and cutting paper stacks for a booklet | 19.8 min |
home_057 |
narrator_F | paper & bookbinding | punching binding holes through a booklet with a guide | 19.7 min |
home_058 |
narrator_B | gardening | planting beans and maize in garden ridges | 19.8 min |
home_059 |
narrator_C | gardening | clearing weeds and mulching in a home garden | 8.0 min |
The full catalogue covers many more home and workplace activities: cooking, repairs, assembly, crafts, cleaning, trades and light industrial work. See Get more data.
What's in each clip
data/
home_001.mp4 video + narration audio (metadata stripped)
home_001_sensors.json IMU samples + device, lens and mount info + per-frame timestamps
home_001_sensors.csv the same IMU samples as flat CSV (stream, t_ns, x, y, z, w)
home_001_frame_timestamps.csv frame_index -> t_ns
home_001_transcript.json word-level ASR transcript (start/end seconds)
home_001_transcript.srt the same, as subtitles
metadata.csv / metadata.json one row per clip
Time base. Every sensor t and frame timestamp is in nanoseconds since the clip's first video frame, on the same device clock. So frame_timestamps.csv maps any video frame to its exact IMU window without resampling.
Narration. Contributors describe what they did and what is around them as they work: which hand, which tool, which object, and where it goes. The narration is useful for video-language alignment, step segmentation and action recognition. Transcripts are automatic and not human-corrected.
How it was made and checked
- Recorded by adult contributors, on their own phones, with our recording app, in their own homes. Contributors signed a recording release assigning the rights in the footage, narration and sensor data, and were paid for accepted recordings.
- Each clip passed our quality review: continuous narration, on-task, stable mount, sensor rate and sync.
- Privacy: every clip was checked frame by frame for faces, other people, screens, documents, names and anything that could locate a home. The narration was checked for names, places and contact details. Clips with any doubt were left out. Two clips are trimmed at the start or end. A few clips have a screen, a printed photo or a background face blurred where needed, and those blurred copies were checked again frame by frame. Every other clip is the original camera file with metadata removed.
- Container and stream metadata (dates, GPS, device serials) were stripped. Contributors appear only as
narrator_Atonarrator_K.
If you believe a clip shows you or contains something that should not be public, email hello@trainthemai.com and we will remove it.
Limitations
- A small sample from 11 contributors in one region. It is not representative of all homes or populations.
- Covers arts, garment making, assembly, home repair, knitting and crochet, gardening, bookbinding, jewellery, food preparation and woodworking. It has no cooking-at-the-stove or cleaning clips.
- ASR transcripts contain errors, especially for object names.
Get more data
Train Them AI collects egocentric video with narration and IMU at scale, in homes and real workplaces across Latin America, Africa and Asia. We can build custom task lists and work in English, Spanish or Portuguese. Licensing can be exclusive or non-exclusive.
hello@trainthemai.com · trainthemai.com
License
Free for research and evaluation under the Train Them AI Evaluation License v1.0. Commercial use needs a separate license. Redistribution is not allowed.
Citation
@misc{ttai_egocentric_home_sample_2026,
title = {Egocentric Home Activities with Narration and IMU (Sample)},
author = {{Train Them AI}},
year = {2026},
url = {https://huggingface.co/datasets/TrainThemAI/Egocentric-Home-Narration-IMU-Sample}
}
- Downloads last month
- 208