Egocentric Last-Mile Package Delivery Video Data

Egocentric last-mile package delivery video data is real-world, first-person video of a courier doing the final leg of a delivery: sorting parcels in a vehicle, carrying one to a door, scanning it and handing it to the recipient. It is recorded from the courier viewpoint, with depth, 3D hand pose, contact events and frame-accurate action segments aligned on one hardware clock.

Delivery differs from warehouse work because it ends with a hand-over to a person. The box and parcel handling page and the logistics fulfillment page cover warehouse, dock and fulfillment work. This page covers the delivery end of the chain, where the parcel leaves the vehicle and reaches a recipient.

The site states delivery in its task catalog, in a Last-Mile Delivery & Courier environment under the Logistics & Warehousing sector. The environment lists four rows: Courier pickup and drop-off, Food delivery, Parcel delivery and Postal sorting. Its examples are parcel delivery, food delivery, courier services, postal sorting and drop-off.

Every row is rated Easy, listed as Available on request and uses the same capture mode, a head-mounted phone or a multiview rig. Two rows describe delivering a parcel to a recipient, one describes delivering a food order, and one is sorting.

The catalog row reads "Sort parcels in the van, carry them to the door, scan and hand them to the recipient." It is the clearest stated last-mile delivery task, because it names the carry to the door and the hand-over to a recipient.

The row says "the van" and "the door", and no more. Vehicle types, package sizes and weights, and door or building access details are Not stated, so they are set in the written spec.

The catalog row reads "Scan and collect parcels at pickup, load the vehicle and hand them over at drop-off." It covers the start of a route, where parcels are collected and loaded, as well as the hand-over at the end.

The row is listed as Available on request, so its timeline is quoted against your spec. The site does not state how many drop-offs a session covers or how long a route runs.

The Food delivery row reads "Collect a food order at the restaurant, bag it in an insulated carrier and deliver it." The site lists it in the same environment, but it describes an order, not a parcel, so this page lists it as neighbouring work and does not count it as package delivery.

It is the only row that names a restaurant or an insulated carrier. The site does not state how the final hand-over of a food order is captured.

The Postal sorting row reads "Sort letters and parcels by route into pigeonholes and trays at the sorting bench." It is sorting at a bench and names no carry or hand-over, so it stays a neighbour.

Warehouse, fulfillment and parcel-handling rows, such as scanning, palletizing, packing and totes, sit in other environments and have their own pages. They do not name a delivery to a recipient or a door.

The catalog places all four rows in the Last-Mile Delivery & Courier environment. The site names no city, neighbourhood, route, carrier or courier, and it does not state capture on a customer own delivery routes or at a customer own facility.

Environment, access and schedule are agreed in the spec. Route lengths and the number of couriers or sites are Not stated.

A head-mounted phone records the first-person view of real work on site, and a multiview rig adds synchronized third-person angles when the spec asks for them. Both are the capture modes the catalog lists for every last-mile row.

The site does not state a camera count, resolution or frame rate for delivery. GPS or location traces are Not stated, so raise them in the spec if your policy needs them.

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 delivery, hand pose tracks the grip on a parcel and the release at hand-over, contact events mark when fingers meet a parcel, and action segments mark each step.

Object labels use the terms your spec sets. Package sizes and weights, force or tactile data, and a delivery-outcome or task-completion label are Not stated. They are defined in the spec if your policy needs them.

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. Delivery episodes use the same schema as every other Firsthand collection, so a loader that reads another Firsthand dataset reads this one without changes.

Every batch is validated against signed accept criteria before it ships, with a reject log, and delivery is rolling, per validated batch.

Every episode links to a signed contributor release through a documented consent chain, and the consent record is part of the delivery. 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 anonymization page states that egocentric capture happens in real homes, workplaces and semi-public spaces, so bystanders end up in frame, and that the pass removes them before anyone downstream sees the footage. The site also states that a clip where a bystander cannot be de-identified is discarded rather than shipped. It does not state a rule for house numbers or addresses, street signs or traffic safety, so this page makes no claim on them.

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. All four last-mile rows are Available on request, so the site states no capture timeline for them beyond a quote against your spec.

You describe the parcels, hand-overs and sequence 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, a reject log ships with every batch, rejected items are recaptured at our cost, and every collection ships under a perpetual, buyer-owned commercial license.