Egocentric Print and Packaging Line Video Data

Egocentric print and packaging line data is real-world video of work at the end of a production line: packing, closing, sealing, labeling, stacking and wrapping. It is recorded against a written spec, with depth, contact events and frame-accurate action segments aligned on one hardware clock where the spec includes them. The tasks it covers are the ones the site states in its task catalog, spread across three environments plus a Print & Packaging Plant environment.

The packing and kitting page covers packing at benches and kitting stations, and the manufacturing workflow page covers whole production procedures. This page covers the packaging-line rows and the print plant environment only.

The site states nine packaging-line task rows. Five sit in a Plastic & Light-Assembly Manufacturer environment, two in a Food & Beverage Manufacturing Plant, and two on a Deformables & Textiles Bench. Eight are rated Easy and one, wrapping and folding, is rated Medium.

The site also lists a Print & Packaging Plant environment, with the examples "Printing presses, die-cutting, folding and packaging lines". That environment holds one row, rated Medium and listed as Ready to deploy, and its description is generic. The table below lists every row with its catalog wording.

Three rows in the Plastic & Light-Assembly Manufacturer environment cover cartons and pallets. The catalog descriptions read "Pack finished goods into cartons at the end of the line and seal and stack them", "Box finished goods, tape cartons shut, stack them on a pallet and stretch-wrap the load" and "Pack products into cartons, build the pallet in layers and wrap it for shipping." All three are rated Easy.

A fourth row reads "Pack finished units into boxes with inserts, seal and label them for dispatch." Carton sizes, pack counts, pallet patterns and line speeds are Not stated.

Two rows in the Food & Beverage Manufacturing Plant cover packs: "Fill packages, run them through the sealer and apply labels before casing" and "Feed packs into the heat sealer, check the seal and run them through the label applicator." Both are rated Easy and listed as Ready to deploy.

A labeling row in the light-assembly environment reads "Apply product and barcode labels to units or cartons by hand or with a label gun." Seal types, film types, label formats and what counts as a good seal are Not stated.

Two rows on the Deformables & Textiles Bench cover wrapping and bagging. "Wrap an item in paper and fold" is rated Medium, and "Bag and seal soft goods" is rated Easy. Both use a multiview rig with handheld grippers and are listed as Available on request.

The wrap row is the closest the site comes to stating folding as a task. It does not describe a folding carton, a box blank or a die-cut sheet. The paper, the item and the fold pattern are Not stated.

Very little, and this page says so. The Print & Packaging Plant environment names "Printing presses, die-cutting, folding and packaging lines" as examples of where capture can happen. Its one row is titled with the environment name and described as "Target environment for commercial egocentric collection."

The site states no task row for press operation, print finishing, die-cutting, folding carton forming or gluing. A print task is therefore scoped in the written spec, not drawn from the catalog.

Packing cartons with void-fill and tape on a warehouse pack bench, and weighing and sealing a package there, sit on the packing and kitting page. Scanning and label printing sit on the barcode scanning and labeling page. Whole production procedures sit on the manufacturing workflow page.

Folding shirts and towels sits on the same bench as the wrap row, but it is a garment task and is not counted here.

The catalog describes steps where the hands change the object as they go: cartons packed then taped shut, a load built in layers then stretch-wrapped, an item wrapped in paper then folded. Film behaviour, paper stiffness, fold sequence and glue handling are Not stated. What the site does state is what egocentric video puts into the training distribution: hands entering the frame from below, occlusion during grasp and motion blur during reach.

RGB alone leaves a model to infer 3D hand structure it cannot see reliably through occlusion and motion blur. Firsthand delivers 3D hand pose as handJoints[2][21][3] in metric units in the head-camera frame, with a visibility flag per joint. Whether it is populated for each packaging row is Not stated.

The catalog uses two modes. For most rows, "a head-mounted phone records the first-person view of real work on site; a multiview rig adds synchronized third-person angles when the spec asks for them." For the bench rows, the mode is "a staged working set recorded by a multiview rig while the operator performs the task with handheld grippers, so demonstrations map onto a robot end effector."

The site states a head camera at 3840 × 2160 and 60 fps for its instrumented rig, but no camera count, resolution or frame rate for the packaging rows, so the capture tier is agreed in the spec.

Each episode follows the Firsthand episode schema: ten streams synced under 2 ms, with head and wrist RGB, depth, 3D hand pose with per-joint visibility, contact and grasp events, and frame-accurate action segments. For packaging work, contact events mark when a hand meets a carton, a film or a label, and action segments mark each step.

Which streams each packaging row populates is Not stated. Force or tactile data is Not stated for these rows unless a spec adds the glove tier, as are success or quality labels for a seal, a label placement or a fold.

The catalog places the rows in a plastic and light-assembly manufacturer, a food and beverage manufacturing plant, a textiles bench and a print and packaging plant. Environment, access and schedule are agreed in the spec, and the number of sites is Not stated.

Every episode links to a signed contributor release through a documented consent chain. Faces, plates and screens are detected and irreversibly blurred at ingest, a reviewer confirms the result before anything ships, and the anonymization record is delivered with the data. An undocumented bystander is a stated reject criterion. How product logos, brand artwork, barcodes and printed labels are handled is Not stated, so name them in the spec.

Delivery formats are RLDS, LeRobot and WebDataset, plus HDF5, zarr and Rerun .rrd. The format table below sets each against what its official documentation states.

Every batch is machine-checked at ingest against the stated reject criteria, reviewed by a person and validated against signed accept criteria before it ships. A rejected item goes to the reject log and is recaptured at our cost. Delivery is rolling, per validated batch.

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.

On-demand top-ups run against the same spec and accept criteria after each training run.

You describe the packaging tasks, the sequence and the capture mode 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.

A Spec Pilot, a small and fast first batch, can prove the spec and accept criteria before you scale. Billing is per validated hour, and every collection ships under a perpetual, buyer-owned commercial license.