Egocentric Fast Food and Quick-Service Kitchen Video Data

Egocentric fast food and quick-service kitchen data is real-world video of work on the line and at the counter of a quick-service kitchen: taking orders, bagging and handing them over, dropping baskets into the fryer, flipping items on the grill and assembling orders in sequence. 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, all in one environment named Fast Food & Quick Service Kitchen.

The food service page covers commercial kitchens and restaurants as a whole, and the meal plating page covers plating and serving to a table. This page covers the three quick-service rows only.

The site states three quick-service task rows, all in the Food & Beverage sector and all in the Fast Food & Quick Service Kitchen environment. Two are rated Easy, counter service and packaging and order assembly, and one, fryer and grill operation, is rated Medium. All three use the same capture mode and all three are listed as Available on request.

The environment examples read "Burger chain, fried chicken, pizza counter, sandwich shop, fast-casual kitchen". These are kinds of operation, not named restaurants, and this page names no chain. The table below lists every row with its catalog wording.

The catalog row reads "Take orders at the counter, bag or box items, add condiments and hand them to customers." It is rated Easy. It is the only quick-service row that includes a customer hand-off, which is why a spec for it should say how the customer appears in the frame and what happens to that footage.

The row does not state a menu, a payment step or a drive-through lane. Order types, bag and box sizes and how long an order takes are Not stated and are settled in the written spec.

The catalog row reads "Drop baskets into the fryer, flip items on the grill, time cooking and drain product." It is the one quick-service row rated Medium. It combines a physical motion, lowering and lifting a basket, with a timing judgment, so the spec has to decide whether the time itself is something the data should carry.

The site states no fryer or grill type, no temperature, no cook time and no food-safety or hygiene procedure for this row, and this page does not add any. Whether a timer is visible in the frame is Not stated.

The catalog row reads "Assemble burgers, wraps and sides from the line in order, wrap and place them on trays." It is rated Easy. It is the closest the site comes to a burger assembly task, and it adds an ordering constraint: the items come from the line in order and finish on a tray.

The row does not state a recipe, an item count or a tray format. What counts as a correct build is Not stated, so the accept criteria are written into the spec.

The site lists a separate Restaurant & Commercial Kitchen environment with rows for line cooking, food prep, dishwashing, line setup and breakdown, portioning and packaging meals, and kitchen deep cleaning. A deep clean row mentions scrubbing fryers, but it is a cleaning task and is not counted as fryer operation here. A Cafe, Bar & Beverage Service environment holds barista work.

Plating dishes and serving to a table sit on the meal plating page, first-person food prep sits on the cooking video page, and the broader commercial kitchen picture sits on the food service page.

The catalog describes steps where the hands change the object as they go: items bagged or boxed, wraps folded around a build, products lifted out of a fryer and drained. Film and paper behaviour, grease, steam and fold sequence are Not stated for these rows. 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 quick-service row is Not stated.

The food service page also states that wet, reflective and hot surfaces get depth-quality flags on affected episodes. That is a program-level statement, not a row-level claim for the fryer row.

The food service page states that failure and recovery cases are part of the same labeled structure as the clean version of a task, and that a written spec can call them out as a requirement for the batch. The site states no failure counts, no failure taxonomy and no success label for the quick-service rows, so any such list is something you name in the spec.

All three rows use one mode: "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." No quick-service row uses the handheld-gripper mode that the bench rows in other environments use.

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 quick-service 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 quick-service work, contact events mark when a hand meets an item, a wrap or a bag, and action segments mark each step.

Which streams each quick-service 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.

The catalog places the rows in a Fast Food & Quick Service Kitchen. 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, and the food service page states that a clip is discarded rather than shipped if someone cannot be properly de-identified, which matters at a counter where customers pass through the frame.

How logos, printed packaging, menu boards and order screens are handled is Not stated beyond the blur rule for screens, so name any branded material 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.

The three quick-service rows are listed as Available on request, which the site defines as scoped and recruited to spec when a buyer asks for it. The site does not state which environments count as deep, so the timeline is quoted against your spec. On-demand top-ups run against the same spec and accept criteria after each training run.

You describe the quick-service 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.