Retail Robotics Training Data

A robot that stocks, faces, or picks on a retail floor needs to learn the same reach and grasp geometry a person uses from that viewpoint, not a description of the task or a fixed-camera view of someone else doing it.

That means synchronized egocentric video, depth, 3D hand pose, and frame-accurate action segments, captured against a named list of shelf-ops skills rather than pulled from incidental footage never meant for manipulation training.

A custom retail collection is scoped to the shelf-ops skills your policy needs: facing and fronting stock to a planogram, case-cut and shelf-load, price and label swap, chilled-case rotation by date code, and hanger or fixture reset.

You describe which of these tasks (and which store formats — grocery aisle, chilled case, apparel fixture, back-of-house staging) your policy needs, and the collection is scoped to that list rather than a generic retail category.

A head- and wrist-mounted camera keeps the facing, restocking, and picking motion in frame from the viewpoint a retail policy will actually see from, across grocery aisle, chilled case, apparel fixture, and back-of-house staging.

Depth, hand pose, and action segments are aligned to video on one hardware clock, so a policy fusing vision with depth or hand pose is training on streams that share the same timeline, not software-timestamped video stitched together after the fact.

Each episode ships with 3D hand pose tracked through reach, grasp, and place; instance segmentation; action segments; and grasp and contact events, alongside the synchronized video and depth streams.

Delivery formats are RLDS, LeRobot, and WebDataset, plus HDF5, zarr, and Rerun .rrd, with readable converter source so you can retarget the schema to your own loader.

Every custom collection ships under a perpetual, buyer-owned commercial license, and exclusivity is available so the data is never resold. Each episode links to a signed contributor release through a documented consent chain.

Because a live retail floor has bystanders and other shoppers in frame, faces, plates, and screens are detected and irreversibly blurred at ingest, and a reviewer confirms the result before anything ships. The anonymization record is delivered with the data.

The existing retail shelf stocking dataset is a fixed, already-captured corpus: a defined hour count, a defined set of skills and store formats, available now under its published terms.

A custom collection is scoped to your written spec instead — the exact skills, store formats, and accept criteria your policy needs, captured on demand rather than chosen from a fixed catalog. If the existing dataset already covers your task list, it ships faster; if it does not, a custom collection fills the gap without waiting on a future dataset release.

You describe the shelf-ops tasks and store formats your policy needs. We turn it into a written spec covering tasks, environments, capture tier, annotation depth, and accept criteria, then quote reach, timeline, and price against that document before any capture begins.

A Spec Pilot is a small, fast first batch that proves the spec and accept criteria before you scale volume on demand, running through the same seven-stage end-to-end pipeline as every Firsthand program: contributor recruiting, capture, per-batch QA, anonymization, and licensed delivery.