Egocentric Garment Production Video Data

Egocentric garment production data is real-world video of a person doing production work while wearing the capture rig: laying out and cutting fabric, feeding pieces through a sewing machine, pressing seams and finished garments, or loading and monitoring mill machines. The footage shows the hands and the material from the head-mounted viewpoint. Each episode follows the Firsthand episode schema, with ten streams synced under 2 ms. Which streams a head-mounted phone populates, and which the rig adds, is confirmed in the spec.

Teams search for it as apparel manufacturing robot training data, a garment production video dataset, egocentric sewing line data and custom textile task data collection. They all describe demonstrations of production work, collected to a written spec. This page covers only what the Firsthand site states and marks everything else as Not stated.

The site states four rows in its commercial task catalog, all in the Textile & Apparel Production environment: "Fabric cutting" (Medium), "Finishing and pressing" (Easy), "Garment sewing" (Hard) and "Textile mill operation" (Medium). All four are Available on request and list the same capture modes.

The first three follow a garment from flat fabric to a finished piece. The fourth is textile work upstream of the garment, since it names yarn and fabric rather than garments. Garment types, fabric types and any step beyond the stated wording are Not stated. A policy that needs another production step writes it into the spec, because a catalog row is a starting vocabulary, not a fixed menu.

The cutting row reads "Lay out fabric, place patterns and cut pieces with shears or a cutting machine", and is rated Medium. The sewing row reads "Feed fabric pieces through an industrial sewing machine to join seams and attach parts", and is rated Hard, the highest rating on the site.

The wording is ordered: lay out, place, cut, then feed, join, attach. This page quotes the tools exactly as the rows name them and adds no others. Machine settings, stitch types, seam counts and cycle times are Not stated.

The pressing row reads "Press seams and finished garments on an industrial iron or press and fold or hang them", and is rated Easy. It closes the garment arc: press, then fold or hang. The mill row reads "Load yarn or fabric onto mill machines, monitor runs and clear faults", and is rated Medium. Monitoring and clearing faults is a different shape of work from handling a single piece, because the wearer watches a running process and steps in when it stops.

Neither row states a garment type, a machine type, a run length or a fault list, and this page does not infer one.

Several other rows touch garments. Tailoring, hemming and custom garment fitting sit in Textile Arts & Stitching Studio, a creative-workshop environment. Folding shirts, fastening buttons and zippers sit in Deformables & Textiles Bench. Folding, hanging and price-tagging apparel sit in a retail floor set. The site describes each as its own environment, and none of them is called garment production.

The table below sets them side by side. For hand and studio sewing, the sewing and textile handling page covers those rows, and for pressing and folding at home, the ironing and garment handling page does.

The environment is listed with the examples "Garment factory, sewing workshop, fabric cutting facility, textile mill, fashion manufacturing", under the Manufacturing & Workshops sector. Those are working production settings, not staged sets. The site states that egocentric capture happens in real kitchens, workplaces and streets.

Whether capture happens at a buyer’s own factory is Not stated, and so are the number of sites, any city or region, and any named facility. Environment, access and schedule are agreed in the written spec, so raise a site-access need before the spec is signed.

In the phone mode, a head-mounted phone records first-person video of real work on site. In the rig mode, a multiview rig adds synchronized third-person angles when the spec asks for them. Every row lists both, and the catalog does not say which applies to which row. Camera count, resolution, frame rate and mounting position are Not stated beyond that.

Each episode follows the Firsthand episode schema: ten streams synced under 2 ms, with RGB, depth, 3D hand pose with per-joint visibility, contact and grasp events, and frame-accurate action segments. The site also states a 200 Hz 6-DoF IMU among its sensors. For garment production, action segments are where a spec marks each step: the layout, the cut, the seam feed, the press and the fold. Contact events mark when a hand meets fabric, a guide or a control.

Not stated: whether depth and hand pose are populated in the phone mode, and whether a result is labeled as a success or quality field. Task duration, the number of episodes or sites, and robot results are Not stated too.

The site gives no technical analysis of garment production, so this page makes no claim about robot results. What the row wording shows is a sequence of steps on soft material, with the material changing shape between steps: flat fabric, cut pieces, a joined seam, a pressed garment. That is a different data shape from picking rigid parts.

The site rates the four rows Easy, Medium, Medium and Hard, which is the only statement on difficulty. It states no success or quality label, so this page makes no claim that any row shows work done correctly, only that each row names a step sequence a spec can capture.

Delivery formats are RLDS, LeRobot and WebDataset, plus HDF5, zarr and Rerun .rrd. Garment production episodes use the same schema as every other Firsthand collection, so a loader that reads another Firsthand dataset reads this one without changes. The format table below sets each format against what its official documentation states.

Every batch is validated against signed accept criteria before it ships, with a reject log, and delivery is rolling, per validated batch. Rejected items are recaptured at our cost, and billing is per validated hour.

Every episode links to a signed contributor release through a documented consent chain, and the consent record is part of the delivery. The release also covers bystanders: contributors confirm they have not knowingly recorded anyone who has not signed it, and where an unidentified person appears, the footage is de-identified or discarded.

A vision model runs on every frame at ingest, flagging faces, heads, licence plates and any legible screen or document. Those regions are irreversibly blurred, and a human reviewer confirms the result before footage ships. If the reviewer is unsure a region is fully covered, the episode is held or discarded, never downgraded and shipped. The site lists no worker-identifying field among delivered fields.

Where coverage is already deep, the first validated batch lands in a median of 48 hours from signed spec. Thin environments typically take three to four weeks, because vetting operators there is the bottleneck, and we quote that up front. All four garment production rows are Available on request, so each is scoped and recruited to your spec first.

The site also states program milestones. A pilot reaches its first delivery in two to three weeks, and a standing program reaches full rate in four to six weeks. Those assume capture coverage is already in place. A batch size and a later-batch turnaround are Not stated, so both are quoted against your spec.

You describe the production steps, garments and materials your policy needs. We turn that into a written spec covering tasks, environments, capture tier, annotation depth and accept criteria, then quote reach, timeline and price before any capture begins. This is the custom egocentric data collection route that every Firsthand task family follows.

A Spec Pilot, a small and fast first batch, can prove the spec and accept criteria before you scale. Every collection ships under a perpetual, buyer-owned commercial license, with a reject log and signed accept criteria on every batch.