Shipping · 1,680 validated hours
Tool use and assembly egocentric dataset
- Episodes
- 11,240
- Participants
- 152
- Capture sites
- 44
- Sync ceiling
- 2.0 ms, hardware-triggered
Related: Custom assembly task data · Egocentric tool use data collection
What a frame contains
Every layer, on every clip in this set.
Rendered from the delivered schema rather than a marketing composite. Toggle layers to see what arrives with the footage.
Synthetic frame, real schema. The sample pack contains the same fields for actual episodes.
Named skills
What the operators were told to do.
Coverage is specified per skill, not per hour. These are the skill lines currently filled in this environment.
- Driver and bit selection, drive to seat
- Clamp, align, and fasten a joint
- Measure and mark to tolerance
- Press-fit and interference assembly
- Wire strip, crimp, and terminate
- Tool hand-off between hands
Environments captured
- Home workshop
- Maker bench
- Small production cell
Condition cells filled
- Daylight
- 77%
- Cluttered
- 83%
- Low light
- 41%
- Failure cases
- 31% — stripped heads, misalignment, rework
Known failure modes
What goes wrong in this environment.
Published because you will find these in the data within an afternoon, and it is cheaper for both of us if you find them in this list first.
- Small fasteners fall below the depth sensor floor at 0.3 m; hand pose remains valid, but metric depth on sub-10 mm parts is not trustworthy and is flagged as such.
What ships with every clip
Ten streams, one clock, one episode file.
Not an à la carte menu. Every validated hour we deliver carries the whole stack, in the schema below, whether you asked for depth or not.
| Stream | Spec | Detail |
|---|---|---|
| Head camera | 3840 × 2160 · 60 fps | Global-shutter, 120° HFOV, rolling-shutter-free, H.265 + lossless keyframes |
| Wrist cameras × 2 | 1920 × 1080 · 60 fps | Left + right, 100° HFOV, rigid mount, extrinsics re-solved per session |
| Depth | 848 × 480 · 30 fps | Active stereo, 0.3–4 m range, metric millimetres, per-frame confidence map |
| IMU | 200 Hz · 6-DoF | Accel + gyro, bias-calibrated, hardware-timestamped on the same clock domain |
| Hand pose | 21 keypoints × 2 hands | 3D metric, per-joint visibility flag, contact events on grasp and release |
| Body pose | 24 joints | 3D, root-relative and world-frame, torso and forearm chains resolved |
| Segmentation | Instance masks | Manipulated objects + target surfaces, tracked IDs across the episode |
| Action segments | Verb + noun taxonomy | 97 verbs, 512 nouns, start/end to the frame, human-reviewed |
| Formats | RLDS · LeRobot · WebDataset | Also HDF5, zarr, and .rrd for Rerun. Converters shipped as source. |
| License | Commercial · buyer-owned | Perpetual, irrevocable, model-weights-clean. Exclusivity available. |
Colour dots map to the modality legend used in every chart on this site. Full field-level schema in the episode schema docs.
Data card
The card that ships with the batch.
Delivered as machine-readable JSON alongside the episodes, so provenance travels with the data instead of living in an email thread.
- Dataset
- Tool use and assembly egocentric dataset
- Validated hours
- 1,680 h — passed sync and QA, shipped to at least one buyer
- Episodes / participants / sites
- 11,240 / 152 / 44
- Sync ceiling
- Δ < 2.0 ms across all streams · median 1.12 ms · p99 1.94 ms
- Consent
- Written commercial release per participant, signed before capture begins
- De-identification
- Faces and licence plates blurred, audio scrubbed, un-blurrable clips discarded
- Collection period
- Rolling. Batch capture dates recorded per episode.
- License
- Perpetual, irrevocable, buyer-owned commercial. Exclusivity available.
- Known limitations
- Small fasteners fall below the depth sensor floor at 0.3 m; hand pose remains valid, but metric depth on sub-10 mm parts is not trustworthy and is flagged as such.
- Formats
- RLDS, LeRobot, WebDataset, HDF5, zarr, Rerun .rrd
Flagged frames are shipped rather than silently dropped. You decide whether to mask, down-weight, or exclude them.
Nearest public dataset
How this compares to EgoDex.
Where a public set is genuinely better at something, we say so. Where the blocker is the license rather than the quality, we say that too.
EgoDex
- Hours
- 829 h
- License
- CC BY-NC-ND 4.0 — no commercial use
EgoDex has genuinely excellent hand pose and is the strongest public reference for bimanual precision. Its license forbids commercial training use and forbids derivatives, so it cannot legally reach your weights. Ours is buyer-owned and perpetual.
Firsthand — Tool use & assembly
- Hours
- 1,680 h
- License
- Commercial, buyer-owned, perpetual
Captured against the named skill list above, hardware-synchronized under 2.0 ms, and shipped with depth, 3D hand pose, body pose, instance masks and action segments on every clip.
See the downstream result →FAQ
Questions about the tool use & assembly set.
The three we get most often on the first call about this environment.
01How many validated hours of tool use & assembly data are available?
1,680 validated hours today, across 11,240 episodes from 152 participants at 44 capture sites. A validated hour is footage that passed cross-stream sync and QA and shipped to at least one buyer — raw recording time is higher. Coverage grows as new condition cells are filled.
02What skills does the tool use & assembly dataset cover?
Coverage is specified per skill against a named spec, not per hour. Current skill lines include Driver and bit selection, drive to seat, Clamp, align, and fasten a joint, Measure and mark to tolerance, and more. Every episode carries frame-accurate action segments so you can filter to the exact skill you are training.
03Can I use the tool use & assembly data to train a commercial model?
Yes. Every set ships under a perpetual, irrevocable, buyer-owned commercial license, with exclusivity available. Each participant signed a written commercial release before capture, faces and license plates are blurred, and un-de-identifiable clips are discarded rather than sold — so the consent chain holds up to legal review.
Pilot tool use & assembly against your spec.
Write the skill list with us, get fifty to a hundred validated hours in two weeks, and read the reject log before you commit to volume.