Egocentric Wiping and Surface Cleaning Video Data

Egocentric wiping and surface cleaning data is first-person video of a person wiping a counter, a table or another surface, clearing it first and resetting it afterwards, recorded from the contributor’s own point of view with synchronized sensor streams alongside. The camera sits on the head, close to what a household robot’s onboard camera would see over a counter or a work surface, so the frames show what the hands and the cloth do rather than a distant view of someone standing near a mess. On this site it is a custom collection: you name the task, the surface, the room and the steps involved, and capture is run to a written skill spec.

The data is human capture. A paid, consented contributor wears the rig and performs the task. It is not robot or teleoperation data, and it is not a fixed corpus you download. This page is narrower than household robot pretraining data, which spans laundry, tidying, cleaning and surface reset across homes, narrower than dishwashing data, which centres on washing, racking and putting away dishes, and narrower than table setting and clearing, which centres on the table. The subject here is the wipe-and-clean-a-surface family of tasks and what the site states about capturing, labeling and checking them.

People reach this topic through close cousins: wiping robot training data, a surface cleaning video dataset, egocentric cleaning task data, or custom household cleaning data collection. They describe one request from different angles, and the work is the same: write down the task, collect it on demand from real counters and rooms, and receive validated episodes. Nothing on this page states that the video carries a cleanliness score, a coverage map or a contact-force reading.

NVIDIA’s GR00T N1 paper describes a data pyramid with web data and human video at the base, synthetic data in the middle and real robot trajectories at the top. First-person human video of wiping and cleaning work sits in that base layer. Human capture does not replace robot data, and this page does not claim that it does.

Wiping looks simple, and a clip that shows a hand somewhere near a counter does not carry what a policy needs. The hand, the cloth, the surface and the moment of contact all have to be present in the episode, and the episode has to say where one step ends and the next begins. That is the difference between footage of someone cleaning and an episode a model can learn a step from.

The site already names this family in several places. Its household chores dataset lists surface wipe and clutter reset among its skills, and describes laundry, tidying, cleaning and surface reset across whole homes. Its kitchen dataset lists wiping and resetting a work surface. Its household pretraining page describes the same cleaning and surface reset work, and the commercial task catalog lists wipe rows across many settings. This page takes that slice and gives it a dedicated spec, so you can ask for the steps your model is missing instead of hoping a broad household collection happens to contain them.

Breadth across surfaces and rooms matters as much as volume on one of them. Generalist’s GEN-0 write-up makes a point about data in different words: "data quality and diversity matters more than sheer volume." For wiping that points at the spread of surfaces, rooms and clutter. The kitchen dataset lists clutter among its conditions, with counters left as found rather than staged, and a counter wiped at the end of a meal looks different from a table cleared in a hotel service set. A corpus that repeats one counter does not cover that spread.

A custom collection lets you aim at the surface, the step and the setting your model is missing. This page does not claim that any particular public corpus lacks wiping or cleaning work.

The first table below reproduces twelve rows of the site’s commercial task catalog that name a wipe or a surface-cleaning step. Each is listed with a skill level, a capture setup and a status exactly as the catalog states them. It is a selection, not the whole catalog. Other rows, such as bathroom and restroom sanitizing with sprayers and mops, kennel cleaning, workshop dust wipe-downs and sand-and-wipe rows, also mention wiping or cleaning, and they are left out because the wipe is one part of a wider task or because they sit in settings this page does not cover.

The three counter and table rows come from the kitchen and hospitality service sets. The catalog lists wipe counter and trash, which is clearing a counter, wiping it and bagging trash, wipe and reset a counter, and clear and bus a table, which is stacking dishes, busing and wiping. All three are listed as easy, use a multiview rig with handheld grippers, and are marked available on request. Banquet plating is a different task that includes wiping plate rims, listed as medium.

The catalog also lists wiping inside other settings. Bathroom cleaning at a hotel, which includes wiping mirrors and surfaces, is marked ready to deploy, and surface sanitizing, which sprays and wipes desks, handles and high-touch surfaces, is marked active collection. Care room, hospital, dental clinic, hair salon, veterinary clinic and gym rows each include a wipe-down step and are marked available on request or active collection as the table shows. These are catalog listings, and the page does not claim more about them than the catalog does. They do show that surfaces, rooms and the cloth-and-spray steps vary a great deal.

The second table shows where the site itself names wiping: the household chores dataset, the kitchen dataset, the in-home and household pretraining pages, and the catalog. It is there so you can see which statements are about homes and which are about other settings, because the catalog rows above are mostly commercial, hospitality and clinical rather than domestic.

No catalog row is titled spill cleanup, and the site does not describe a cloth type, a sponge, a surface material, a stain type or an area size for any of these rows, so none is stated here. A surface, stain, spill or setting outside these rows is a skill-spec question to settle in a Spec Pilot rather than an assumption. The catalog is a list of what is captured or available, not a ceiling on what can be briefed.

A wiping collection uses the same capture stack as every Firsthand program. The instrumented tier is a head-mounted stereo rig that records first-person RGB, with a metric depth stream recovered from the calibrated stereo rig rather than from 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 second, closer view of the hands than the head camera alone, which matters when a hand presses a cloth flat or reaches into a corner.

Every stream is hardware-triggered onto one PTP clock domain, so depth, motion and hand pose line up with the same video frame rather than being approximated afterwards. The rig is calibrated twice per session, before capture starts and after it ends, and an episode is accepted only if RMS reprojection stays under 0.28 px. The catalog lists the capture setup per task: a multiview rig with handheld grippers for the counter, table and care-room rows, and a head-mounted phone or multiview rig for the bathroom, facility, clinic, salon, gym and banquet rows. Which tier applies to your program is set in the skill spec.

On depth, the site states that wet and reflective surfaces degrade active-stereo depth, which is why per-frame confidence maps are shipped so you can mask rather than guess. A wiped counter is wet and glossy for a moment, so that statement applies directly. The site does not state how much depth is lost on a particular surface, a cloth or a film of water, and it does not state a camera count, resolution or frame rate specific to wiping. Force and tactile readings are not among the streams, and no robot or teleoperation data is captured.

Each attempt is delivered as one episode. Inside it, action segments divide the episode into start- and end-timestamped spans, each carrying start_ns, end_ns, verb and noun. For this task family a segment can be one step as written in the skill spec, so a counter reset becomes a sequence of labeled steps such as clear, wipe and put back rather than one long clip. The other standard fields are hand_joints and hand_visible for 3D hand pose, contact_events for the moment the hand meets an object, objects, and instance_masks, stored at the depth resolution of 848×480 so they line up with the depth stream.

Hand pose is delivered for both hands with a per-joint visibility flag for occluded fingers. That is the label that carries the two-handed structure of wiping: where one hand steadies or lifts an item while the other works the cloth. Action segments mark where one step ends and the next begins. The site describes contact events as the moment the hand meets the item, and does not say whether a cloth or the surface itself counts as the item, so the spec is the place to settle what a contact event means for your task.

Masks need a spec decision. The site states that clothing, bedding and bags do not track well with masks propagated frame to frame, so deformable captures ship per-frame instance masks where the spec calls for them. Cloths, sponges and rags are not named in that statement. Whether a wiping cloth needs per-frame masks is settled in the spec rather than assumed.

The site does not list labels for how clean a surface became or how the cloth moved. There is no stated label for a coverage map, a cleanliness score, a stain or residue annotation, a wipe pattern, a contact force or pressure reading, or a task-completion score. Several catalog task descriptions mention checking, disinfecting or resetting, but that describes the task the contributor performs, not a delivered measurement.

Several limits the site already states apply directly to wiping capture, and several it leaves open. Each stated limit below appears elsewhere on the site rather than being introduced here.

Hand occlusion is the first. Fingers disappear behind objects in manipulation, which is why 3D hand pose carries a per-joint visibility flag. A hand pressed flat under a cloth is an everyday case here. The second is wet and reflective surfaces, which degrade active-stereo depth and are the reason confidence maps ship. The third is masks, where per-frame instance masks are the stated answer for items that do not track well when propagated, and the spec is where you decide whether a cloth needs them.

The fourth is the set of machine-checked reject criteria. An episode is rejected for a sync breach beyond the ceiling for more than three consecutive frames, calibration drift beyond tolerance, the manipulated object out of frame for more than 15 percent of the episode, more than 35 percent invalid depth pixels in the manipulation region, an undocumented bystander, or a mismatch against the named skill spec. A human reviewer then checks each episode against the task spec.

The kitchen dataset also names the conditions that stress a household capture: clutter, with counters left as found, and failure cases such as drops, spills, mis-grasps and recoveries. Those are listed as conditions of that dataset, and spills there are a failure case rather than a cleanup task. What the site does not state is just as useful to know before you scope. It gives no mask accuracy figure for cloths, no description of how glare on a clean counter or steam above a wet one is handled for wiping, and no cleanliness or coverage annotation. If your policy depends on any of those, say so at the start.

Validation runs against signed accept criteria written into the spec before capture. Each batch goes through per-batch QA and a reject log ships with it. Only footage that clears the criteria is billed as a validated hour, and rejected items are recaptured at our cost. Many teams start with a Spec Pilot, a small, fast first batch that proves the spec and the accept criteria on real footage before volume scales.

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, and speech is scrubbed of names, addresses and card numbers while ambient audio is kept. A human reviewer confirms that no un-consented person is identifiable in any frame, and footage that cannot be de-identified is discarded. The anonymization record is delivered with the data.

Homes add their own conditions. The site states that household capture happens inside lived-in spaces, that bystanders are de-identified or the footage is discarded, and that bathroom capture excludes any episode with a person in frame other than the consenting operator. A counter in a home can carry mail, papers or other printed items. The site states that mail and barcodes are not listed as blur categories, so this page does not claim they are blurred. The listed categories are faces, plates and screens. If text on items in frame matters for your use, raise it in the spec so the handling is agreed before capture rather than assumed afterwards.

On safety and qualification, the site states no contributor safety protocol, no chemical-safety guidance and no safety certification for any wiping or cleaning row, and this page does not describe one. If your program needs terms about who performs the work, the products used or the conditions under which it is done, raise them when you scope it. None of this is legal advice, and the executed agreement controls.

From signed spec to first validated batch, the timeline is 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. Speed comes from a warm pipeline, not from cutting consent, review or anonymization. After the first batch, on-demand top-ups run against the same accept criteria, so you can add more surfaces, rooms or variation without rewriting the spec. The site does not state a turnaround for those later batches, so plan them in your quote.

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. Collections are billed per validated hour.

The license is perpetual, irrevocable and worldwide, with the data buyer-owned and non-exclusive by default. Training, evaluating, fine-tuning and deploying models, including commercial ones, is permitted, and exclusivity is available as a separate term. The executed agreement controls. The site does not publish a price list, a minimum order or a batch size for this task family, so those are set in your quote.

Where this site does not state something, this page says "not stated" rather than filling the gap. The table below separates what is delivered from what is not stated. There is no listed label for coverage, cleanliness, stains or wipe pattern, no force or tactile stream, no cloth, surface or product specification, and no area size. Nothing here claims a training outcome or a model improvement.

The tradeoff to plan for is the embodiment gap between a human hand and a robot end-effector, which retargeting or pairing with robot data has to close. Counts of homes, contributors, hours and tasks are set in your skill spec.

Describe the surfaces, the steps, the rooms 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. The scoping checklist below maps the parts of a skill spec onto a wiping request. A published sample recording is on the sample data outputs page if you want to see the streams before you write a spec.

Three choices matter most for this family. Name the step and the surface rather than a broad activity, because that is what lets each step become one action segment. Name the setting and the capture tier, since a home counter, a hotel bathroom and a clinic room present different geometry and the catalog lists a different capture setup for them. And name the lighting and participant mix: daylight coverage fills fastest, while evening, low-light capture is the harder cell to fill.