Egocentric Pallet Building and Wrapping Video Data

It is first-person video of a person working a pallet: stacking boxes into a stable pattern, taking a load apart, or wrapping it in stretch film. The camera is head-mounted, with two wrist cameras for a closer view of the hands. The contributor is a paid, consented person following a written skill spec, not an actor and not a robot.

A pallet task is long and repetitive: many placements add up to one load. The value of the data is what the hands do at each step, labeled so a single placement or wrap pass can be sliced out. Where the site states nothing, a table below says "Not stated".

The task catalog already names pallet work. Eight rows name a pallet, and this page reproduces them as the catalog states them.

Three sit in the Warehouse Zone, whose examples read "Conveyor, racking, pallets, totes, a pack bench and scale." "Palletize boxes" is described as "Stack boxes in a stable pallet pattern" and "Stretch-wrap pallet" as "Wrap a pallet with stretch film", both listed as medium skill, and "Depalletize to shelf" is listed as easy. All three use the capture setup "Multiview rig + handheld grippers" and are marked available on request.

Three more sit in the Warehouse & Distribution Center, whose examples read "E-commerce fulfillment hub, distribution center, cold storage facility, freight warehouse." They are the outbound freight row, "Loading Outbound Trucks" and the inbound pallet receiving row, with their descriptions in the table below. All three are listed as medium skill, use "Phone (head-mounted) / multiview rig" and are marked ready to deploy.

Two more sit in the Plastic & Light-Assembly Manufacturer, in the Manufacturing & Workshops sector: "Box and palletize finished goods" and "Boxing & Palletizing Finished Goods", described in the table below. Both are listed as easy, use the same phone or multiview capture and are marked ready to deploy.

Two rows state equipment, a hand truck and a manual pallet jack, repeated here only as listed. The site states no forklift or other vehicle, so none is claimed, and it lists no row for labeling a pallet or strapping a load. Pallet patterns, layer counts, load weights, wrap counts, film types and case or SKU counts are not stated, so they are decisions for the written spec.

An episode is one contributor performing one task from the skill spec, such as building one pallet or wrapping one load. The wider stack is ten streams, including a head-mounted global-shutter 4K/60 camera with a 120 degree lens, two wrist-mounted 1080p/60 cameras, depth, 200 Hz 6-DoF IMUs, 3D hand and body pose, instance segmentation and action labels, hardware-aligned within the 2 ms ceiling the site documents.

Every episode carries int64 nanosecond timestamps per stream, camera intrinsics and extrinsics, hand joints for both hands, instance masks, action segments and contact events. The rig is calibrated at the start and end of every session, and an episode is accepted only if RMS reprojection stays under 0.28 px.

The site states no field for pallet pattern, layer count or load state, no stack-stability label, and no load weight, wrap count or film type.

The unit of labeling is the action segment. The site states a taxonomy of 97 verbs and 512 nouns, annotated start and end to the frame and human-reviewed. A spec that says "palletize the boxes" gives you one long segment. A spec that names pick up, place and square the box, then start the next layer, gives a sequence you can slice by placement.

Hand pose carries a per-joint visibility flag, contact events mark when a hand meets the item, and instance masks are stored where the spec calls for them.

Data is delivered in RLDS, LeRobot, WebDataset, HDF5, zarr or Rerun .rrd, with converters shipped as readable source. You receive a perpetual, irrevocable, buyer-owned commercial license, the consent artefacts for the batch and a data card that lists jurisdictions of capture. Billing is per validated hour, and rejected episodes and recaptures are never billed. The site states no stack-stability, wrap-quality or other success label.

The site states no pallet learning result, so this is plain reasoning about the task, not a finding. A pallet is built from many placements that depend on one another, so one clip holds a long sequence.

Three consequences show up in the data. First, sequence: the order of placements is part of the task, so action segments let one placement be sliced out and kept in order. Second, occlusion: a box or a stacked layer hides fingers, so the visibility flag marks joints that were not seen instead of presenting a guess. Third, wrapping is a continuous motion around the load, and the spec decides how it is split into segments.

Firsthand captures people, not robots. There is no robot joint-angle log, robot result, benchmark or teleoperation data for these tasks, and using the episodes for a specific robot means your own retargeting step. Nothing here claims a training outcome.

The standard capture lists no force or tactile stream, only contact events, and the site states no label for load weight or grip force. A separate glove tier is described on its own page, and this page makes no claim about it for pallet work.

The catalog places the pallet rows in a warehouse and distribution center, a warehouse zone with conveyor, racking, pallets and a pack bench, and a plastic and light-assembly manufacturer. The site does not state which sites or countries host capture, or how many, and it states no capture at a buyer’s own warehouse or facility. The spec settles location and access.

Only what the site already states applies to privacy. Every participant signs a written release granting commercial training rights before capture begins. Faces and licence plates are detected and irreversibly blurred, a human reviewer confirms that no un-consented person is identifiable in any frame, audio is scrubbed of identifying speech, and a clip where a bystander cannot be de-identified is discarded rather than shipped. This is not legal advice, and the executed agreement controls.

The site states one published timeline: from a signed spec to a first validated batch, a median of 48 hours where coverage is already deep, and typically three to four weeks where coverage is thin because operator vetting is the bottleneck.

Five of the eight pallet rows are marked ready to deploy and three available on request. The site does not say which timeline applies to which, so ask when you request the quote.

Start with the free, ungated sample pack to check the loader and the label quality, then move to a Spec Pilot: a small first batch against the written spec that proves the fill rate and the reject rate on your conditions before you scale.

A good brief names the pallet task, the order of steps, how each step starts and ends, and what counts as a reject. The site’s skill-spec guide says to name the verb and the noun rather than a broad activity. Then request a custom dataset.