Egocentric Mural Painting Data

It is real-world, first-person video of a person doing wall painting work, recorded with a head-mounted phone or a multiview rig, together with the streams the site already lists for an egocentric episode. The catalog describes the work in its own words: sketch the outline on the wall, then paint large areas and details with rollers and brushes. The buyer sets the task and the accept criteria; Firsthand handles contributors, capture, QA, consent and delivery.

The data is human capture by paid, consented contributors, not robot or teleoperation data. This page is about what a policy can learn from watching wall painting work, and it makes no claim that any robot has been trained on it or that any robot paints a wall. People reach this topic through several phrasings, such as first-person wall art video, a large-surface brush and roller dataset or custom art studio data collection. They describe one request from different angles.

One. In the Art Studio (painting, sculpture, printmaking) environment of the Creative Workshops & Crafts sector the catalog lists "Mural work", described as "Sketch the outline on the wall and paint large areas and details with rollers and brushes." It is rated Medium, lists the capture setup Phone (head-mounted) / multiview rig, and has the status Available on request. The environment examples read "Artist studio, canvas painting, sculpting, mural work, printmaking". The first table below reproduces the row exactly.

Available on request means the work is scoped and recruited to spec when a buyer asks for it. This page therefore claims no completed collection, no delivered hours and no delivered footage for the row. The site names no mural, no wall, no artist and no studio, and no off-the-shelf clip shows this task.

Three other rows share the Art Studio environment: "Canvas painting" (Medium), "Printmaking" (Hard) and "Sculpting" (Hard), all Available on request with the same capture setup. They are context only on this page. Canvas painting is the nearest neighbour: it mixes paint on a palette and applies it to canvas with brushes and knives, so it differs from the mural row in surface and in tools, a canvas against a wall, and a knife alongside brushes and rollers.

Painting of building surfaces is a separate topic. The task catalog lists "Painting & Priming" and "Prime and paint interior walls using rollers and brushes" at an active construction site, "Roll + cut-in paint" at a trades workshop bench and "Staining & Finishing" at a carpentry workshop, as trade tasks with their own statuses. This page covers the art studio row only.

Not stated. The catalog lists a head-mounted phone or a multiview rig for the row and no stereo pair. Stereo applies here only through the site general stereo offer: the site describes a stereo head rig and a hard-environments stack, and a stereo pair for a mural order is a choice made in the skill spec and confirmed on a Spec Pilot.

This is why the page is written as an egocentric page for one row rather than as a description of existing stereo footage. If you need stereo, say so in the brief and see the stereo art studio page, which covers the same row alongside canvas painting, printmaking and sculpting with an MCAP delivery focus. If you are happy with the catalog setup, the egocentric route described here is the one the row lists.

This page is egocentric capture of the one Mural work row. The egocentric arts and crafts page covers all four Art Studio rows as a family and treats the mural row as one of four. The stereo art studio painting and sculpting page covers the same four rows with stereo and MCAP as its subject, so mural work is one row among four there too.

The egocentric painting and coating page is about trade painting, and mentions the two art studio painting rows only as artistic listings. The stereo painting and coating page is stereo, but it counts four finishing-trade rows and leaves canvas and mural work out. The second table below puts them side by side with their status as the site lists it. No page on the site has an art studio mural row as its main topic other than this one.

The catalog words the row as a short ordered job: sketch the outline on the wall, paint large areas, then paint details, with rollers and brushes. That gives three stages a spec can name. The first is a line drawn on a surface, the second is wide coverage with a larger tool, and the third is closer work with a finer one.

What a spec can do is cut the job into steps the buyer names and give each a start and an end. Action segments carry start_ns, end_ns, verb and noun, so one session becomes a sequence of labeled spans instead of one long clip, and contact events mark when a hand meets a tool or the wall. The site names no stroke, no coat, no pass count and no finishing step, and it lists no label for whether a mural is finished, so a spec that needs them sets them. How many episodes one wall becomes, and whether an episode is a single step or a run of work, is Not stated.

In the phone mode, a head-mounted phone records first-person video of real work on site. In the rig mode, a multiview rig adds synchronized third-person angles when the spec asks for them. The row lists one setup that covers both, so which applies to a batch is settled in the spec. Camera count, resolution, frame rate and mounting position are Not stated beyond that.

For a large surface, third-person angles can show where the wearer stands relative to the wall, which a head view alone does not, and a spec that wants them asks for the rig mode. The site states that egocentric capture happens in real kitchens, workplaces and streets, and the environment examples describe a working artist studio. Whether capture runs at a buyer wall, an artist studio or another place is Not stated, and no site is named. Access and schedule are agreed in the written spec.

Each episode follows the Firsthand episode schema: ten streams synced under 2 ms, with RGB, depth, 3D hand pose with per-joint visibility, contact and grasp events, and frame-accurate action segments. The site also states a 200 Hz 6-DoF IMU among its sensors, and the language data is one task instruction per episode. Calibration is measured every session and shipped with the episode.

Not stated: whether depth and hand pose are populated in the phone mode, a success or quality field for a result, task duration, the number of episodes or sites, and robot results. This page claims no affordance labels, post-contact trajectories, reward labels, robot trajectories or paired human-robot video for the row.

Hand occlusion is the first case. A roller or a brush covers the fingers that hold it, and a closing hand hides joints. The site states that the hands occlude the work and that fingers disappear behind objects. The per-joint visibility flag marks joints that are not visible instead of guessing them, contact events mark when a hand meets the tool or the wall, and a spec can ask for multiview angles so a grip the head camera loses is still recorded.

Large-surface head motion is the second. A mural asks the wearer to move along a wall, and the sensors include an IMU that records that motion, but the site states no head-motion figure for the row. Thin, low-contrast strokes are the third: the site says metric depth on sub-centimetre parts is flagged rather than interpolated, and whether a fine detail line counts as such a part is Not stated, so a Spec Pilot checks it on real footage.

Wet and reflective surfaces are the fourth. Fresh paint is a wet surface, and the site states that depth degrades on wet and reflective surfaces, which per-frame confidence maps handle by masking low-confidence regions instead of hiding them. Lighting is the last: the site states no standard lighting values for the row, so conditions are fixed in the skill spec for the place in question.

Every batch is validated against signed accept criteria before it ships, with a reject log. The site states that an episode is rejected when the manipulated object is out of frame for more than 15 percent of it, when more than 35 percent of depth pixels in the manipulation region are invalid, when a sync breach runs beyond the ceiling for more than three consecutive frames, when an undocumented bystander appears, or when the episode does not match the named skill spec. Failed episodes are rejected and logged per batch, not delivered with a caveat.

For a wall, what counts as the manipulated object, the tool, the painted area or the wall itself, is a spec decision, and it decides how the out-of-frame rule applies. Billing is per validated hour, and rejected items are recaptured at our cost. The site does not state an accuracy figure for step labels, and it lists no label for the quality of a painted wall.

The site off-the-shelf catalog lists about 324,800 hours across 18 datasets. None of its dataset descriptions names mural work or wall painting, so no off-the-shelf set stands in for this row. The published sample clips are format examples, not coverage of any art studio row.

Mural collection to a spec is therefore custom collection: custom collection at scale through a network of thousands of capture providers, for a consistent, recurring data flow to labs. Where coverage is deep, the median time from signed spec to first validated batch is a median of 48 hours; where it is thin, turnaround runs three to four weeks. The row is Available on request, so the scoped-to-spec timeline applies rather than the ready-now case, and the quote states it up front.

Every episode links to a signed contributor release through a documented consent chain, and the consent record is part of the delivery. Faces are blurred at ingest, confirmed by human review, with the anonymization record delivered alongside. Footage that cannot be de-identified is discarded, and an undocumented bystander is a stated reject criterion. An existing artwork, a mural design or a signature in frame is not listed as a blur category, so raise it in the spec if it matters.

The buyer owns the dataset under a perpetual, irrevocable, non-exclusive-by-default licence, with exclusivity available. Consent is recorded per session. The executed agreement controls, and the anonymization page describes the steps in more detail.

Paint types, roller and brush models or counts, wall or surface types and sizes, artworks, piece counts, task times, error rates, force or tactile data, a quality or completion label and any safety or compliance certification are not stated on the site as standard fields. No artist, studio, client or brand is named for the row.

If your policy needs any of these, they are written into the skill spec as annotation depth and covered in the quote before capture begins. This page gives no painting instructions and no advice on ladders, scaffolds, heights or other equipment or workplace safety.

Describe the row, the steps you expect to see and where each starts and ends. Say whether one episode is a single step, a single wall or a run of work, what counts as the object, what counts as a reject, and whether you want the head-mounted phone, a multiview rig or a stereo pair through the general stereo offer. Then name the streams you want and the delivery format.

Most programs begin with a Spec Pilot, a small first batch that proves the spec and accept criteria on real footage before you scale. The last table below shows what each spec element decides.