Egocentric Price Tagging and Shelf Signage Data

Egocentric price tagging and shelf signage data is first-person video of one person printing price labels, tagging merchandise with a pricing gun and placing shelf signage, recorded from that person’s own point of view. A paid, consented contributor wears the capture rig and performs a task written in a skill spec, and each attempt is delivered as an episode with synchronized streams, hand pose and labeled action segments. It is human capture. It is not robot or teleoperation data, and it is not a fixed corpus you download.

People search for this under several names: price tagging robot training data, a shelf signage video dataset, egocentric retail associate data or custom retail task data collection. They describe one request from different angles. The work is the same: write the task down, collect it on demand, and receive validated episodes.

The catalog lists three rows that name a price tag, a price label, a pricing gun or shelf signage. Price Tagging & Signage reads: "Print price labels, tag merchandise with a pricing gun and place shelf signage." Tag and price merchandise using a pricing gun and label sheets reads: "Load the pricing gun, tag each item and apply labels from sheets to packaging." Price-tag items reads: "Attach price tags/labels." The first table below reproduces all three exactly as listed.

Both back-of-house rows are rated Easy, captured with a head-mounted phone or a multiview rig, and marked ready to deploy. Price-tag items is rated Easy, captured with a multiview rig and handheld grippers, and marked available on request. Other retail rows mention labels or tags in passing, and the second table sets them aside so the boundary is visible.

The environment examples for Retail Store -- Back of House read: "Stockroom, receiving dock and backroom of supermarkets, apparel and general retail." Both rows are set there, so a spec should name the merchandise type, the tool, whether labels come from sheets or are printed first, and where one tagged item ends and the next begins.

Price Tagging & Signage spans three actions in one row: printing labels, tagging with a pricing gun and placing shelf signage. The spec should say whether an episode is one action, a run of items or the whole sequence. The site does not state a merchandise type, a tag or sign size, a gun model or a label printer, so those are yours to set.

Price-tag items sits in Retail Floor Set (shelving, checkout, rail, pack bench), described in the catalog as "Shelving, a checkout, a garment rail and a pack bench." It is captured with a multiview rig while the operator uses handheld grippers, so demonstrations map onto a robot end effector. That differs from the head-mounted capture of the back-of-house rows, so say which one the program wants.

The row text is only "Attach price tags/labels." The site does not say what the tags are attached to or how many. Neighbouring floor-set rows include Fold + stack apparel and Hang + size-order rail, but this page covers only the price-tag row.

Each attempt is delivered as one episode. The site does not set an episode length or count, so the skill spec does. Inside the episode, ten streams run on one hardware clock, and action segments divide the recording into spans, each carrying start_ns, end_ns, verb and noun. A tagging run can become labeled steps, such as loading the gun, tagging an item and applying a label, as the spec defines them, rather than one long clip.

Two machine-checked rules apply. An episode is rejected if the manipulated object is out of frame for more than 15 percent of it, and if more than 35 percent of depth pixels in the manipulation region are invalid. A tag or label is small, so the spec should say which object counts as manipulated at each step.

The instrumented tier is a head-mounted stereo rig that records first-person RGB, with metric depth recovered from the calibrated stereo rig rather than a standalone depth sensor, and 6-DoF IMU data at 200 Hz from head- and wrist-mounted sensors. The wider stack is ten streams: a 4K head camera, two wrist cameras, depth, IMU, 3D hand and body pose, instance segmentation and action labels, hardware-aligned under 2 ms. The two wrist cameras give a closer view of the hands than the head camera alone, which matters when a hand squeezes a pricing gun.

The rig is calibrated twice per session, and an episode is accepted only if RMS reprojection stays under 0.28 px. The site states that small fasteners can fall below a depth sensor’s reliable floor, so sub-centimetre depth is flagged rather than interpolated, while hand pose stays valid at that range. It does not state how depth behaves on a flat adhesive label, a price tag or a printed sign, so this page does not either.

Each episode carries hand_joints and hand_visible for 3D hand pose, contact_events for the moment the hand meets an item, objects, and instance_masks stored at the depth resolution of 848×480. Hand pose is delivered for both hands with a per-joint visibility flag for occluded fingers, and contact events mark when a hand meets the merchandise, the gun or a sign.

The site does not list labels for what a tag or sign says. There is no stated label for a price, a product, a read tag, a correctly placed label, a legible sign or a pricing-accuracy result. A task description that mentions printing a label describes what the contributor does, not a delivered measurement.

Every contributor signs a written release before capture, and each delivered episode links to that record through consent artefact IDs in the batch data card. The consent chain has six checkpoints, from before capture to after delivery, and a break at any one discards the footage rather than shipping it at a lower grade. At ingest, automated detection finds faces, plates and screens and blurs them irreversibly. A human reviewer confirms that no un-consented person is identifiable in any frame, and footage that cannot be de-identified is discarded. There are no participants under 18.

Two uses of the word signage should not be confused. Bystander signage is the notice the site states is posted where a capture site is semi-public. Shelf signage is the task. The site does not state that readable signs, price labels or tag text are blurred, so raise any such need before the spec is signed. None of this is legal advice, and the executed agreement controls.

The catalog names two settings: Retail Store -- Back of House and Retail Floor Set (shelving, checkout, rail, pack bench). The site names no store, chain, retailer, point-of-sale system or city, and it lists capture at your own facility as not offered. This page therefore makes no claim that episodes are recorded in a particular store, and none that they are recorded at yours. Capture is run to your skill spec using the settings and contributors Firsthand arranges.

Delivery is in the format your stack already reads: RLDS, LeRobot, WebDataset, HDF5, zarr or Rerun .rrd, or raw streams with metadata. Readable converter source ships alongside the data so the schema can be retargeted to your own loader. The official documentation for three of these formats is linked in the sources below.

Collections are billed per validated hour, and only footage that clears the signed accept criteria is billed. Rejected items are recaptured at our cost. The license is perpetual, irrevocable and worldwide, buyer-owned and non-exclusive by default, with exclusivity available as a separate term. The site publishes no price list, minimum order or batch size for these rows, so those are set in your quote.

From signed spec to first validated batch, the site states a median of 48 hours where coverage is deep and three to four weeks where it is thin and contributors need to be recruited first. Two of the three rows are marked ready to deploy, which the site defines as contributors and site access already set up, with capture starting once a spec is signed. The site does not say whether coverage for these rows counts as deep or thin, so ask which timeline applies when you request the quote.

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. It does not state a turnaround for later top-ups, so plan them in your quote.

The barcode scanning and labeling page covers scanner, shipping-label and parcel-label work across six sectors and lists price tagging at a retail back of house as one row among about twenty. This page takes the three price-tagging rows and treats them in depth. The retail robotics training data page covers shelf stocking, restocking, facing and picking, which this page does not.

The grocery put-away and grocery bagging pages cover separate rows in the grocery setting. None of the neighbouring pages is a substitute for the price-tagging rows listed here.

Where this site does not state something, this page says so. There is no listed row for shelf-edge label swaps, sale-sign changeovers or price auditing. There is no tag or sign size, merchandise type, gun model, label printer, point-of-sale system, store or chain, region, episode count, or success or quality label. There is no pricing-accuracy claim anywhere on this page, and nothing here reports a robot result, a benchmark or a model improvement.

Counts of contributors, hours and tasks are set in your skill spec. A task outside the three rows is settled in a Spec Pilot rather than assumed.

Describe the task, the items, the tool, the setting and the variation or failure cases your pipeline is missing. That becomes a written skill spec, quoted for reach, timeline and price before any capture begins, and most programs run a Spec Pilot first, a small, fast first batch that proves the spec and the accept criteria on real footage before volume scales. A published sample recording is on the sample data outputs page if you want to see the streams first.

Name the step and the item rather than a broad activity such as "retail work," because that is what lets each step become one action segment. If the program needs shelf signage placement specifically, say so at the start, since it is one of three actions inside a single row.