Comparisons

Compare egocentric data, honestly.

How Firsthand stacks up against the public datasets teams reach for first, what their licenses actually allow, and a neutral checklist for evaluating any vendor. Third-party facts are marked for verification and sourced.

At a glance

Firsthand vs the public datasets.

The dimensions that decide most robot-learning projects. Full breakdowns on each comparison page.

DimensionFirsthandEgo4DEgoDexEgoScale
Hand pose3D, handJoints[2][21][3], metric, in the head-camera frameNo manipulation-grade 3D hand poseStrong 3D hand poseNot manipulation-grade 3D hand pose throughout [VERIFY]
Action labelsFrame-accurate action segments (verb, noun, ns timestamps)Dense narrations and task benchmarks; not frame-accurate manipulation segments throughout [VERIFY]Task/annotation scheme per the dataset release [VERIFY]Scaling-study labeling [VERIFY]
SyncHardware-triggered; median 1.12 ms, 2.0 ms ceiling, drift written to metadataNot hardware-synchronized to a sub-2 ms ceiling [VERIFY]Not specified as sub-2 ms hardware sync [VERIFY]Not a sub-2 ms hardware-sync guarantee [VERIFY]
Commercial use allowedYesPermitted under the Ego4D license terms [VERIFY]No — non-commercial, no-derivatives[VERIFY]
CostPer validated hour, quoted per engagement; free sample pack, no formFree under license termsFree for non-commercial researchPer the release [VERIFY]

Figures marked [VERIFY] are confirmed against each dataset's official sources before publishing.

Compare it on your own data.

Get the free 40-episode sample pack and score it against any dataset on the checklist.