Now recruiting capture teams worldwide. Get paid to record first-person video of real work.

Apply to capture

Task family

Egocentric Concrete Finishing Data

Concrete finishing is an ordered run of two-handed, hand-tool steps over a surface that changes under the tool: shovel the mix into the form, screed it, float it, trowel it. This page covers the egocentric side: first-person screed and trowel video and a concrete pour and float dataset, recorded from the contributor’s own point of view. We run custom construction data collection fast and on demand against a spec you write, with custom collection at scale through a network of thousands of capture providers, for a consistent, recurring data flow to labs.
Concrete row
"Concrete work": Hard, Available on request
Catalog capture setup
Phone (head-mounted) / multiview rig
Streams
Ten, synced under 2 ms
License
Perpetual, buyer-owned

Requirements

What does a first-person concrete finishing clip need to show?

01

Both hands, the tool and the surface in view

The concrete row names shovels, floats and trowels. First-person capture keeps both hands, the tool in the hand and the surface in one frame, and 3D hand pose carries a per-joint visibility flag, so a finger hidden behind a handle is flagged rather than guessed.

02

A spec that names the step

Name the verb and the noun, such as mixing, pouring, screeding or finishing, and where each pass starts and ends, so each pass becomes a labeled action segment instead of one long clip.

03

Contact tracked through the pass

Hand pose for both hands, with contact events marking when a hand meets the tool or the item, carries the sequence that a clip of someone standing near a pour does not.

04

Reject rules written before capture

An item is rejected when the manipulated object is out of frame for too much of the episode, when depth is invalid over too much of the manipulation region, or when an undocumented bystander appears. The thresholds are agreed in the spec.

Definition

What is egocentric concrete finishing data?

Egocentric concrete finishing data is first-person video of a person doing concrete work, recorded from the contributor’s own point of view with synchronized sensor streams alongside. The task is the one the site’s catalog lists: mix, pour, screed and finish concrete into forms using shovels, floats and trowels. The head-mounted camera sits close to what a robot’s onboard camera would see, so the frames show what the hands and the tool are doing to the material rather than a distant view of someone standing near a pour.

On this site it is a custom collection: you name the task and the setting, and capture runs to a written skill spec. The data is human capture by paid, consented contributors, not robot or teleoperation data, and it is not a fixed corpus you download. The same request is also searched as a concrete pour and float dataset, first-person screed and trowel video or custom construction data collection.

Catalog row

Which catalog row covers concrete finishing?

One row is a concrete finishing row, reproduced exactly in the first table below. "Concrete work" sits in the Heavy Industry & Fabrication environment, in the Manufacturing & Workshops sector, and is described as "Mix, pour, screed and finish concrete into forms using shovels, floats and trowels." It is rated Hard, lists "Phone (head-mounted) / multiview rig" as its capture setup, and its status is Available on request.

A second row, "Construction General Skilled Trades", sits in Active Construction Site, in the Construction & Trades sector: "Skilled trade work on active sites including framing, concrete and structural tasks." It is rated Hard and Ready to deploy. It is a general trades row that names concrete among other work, not a concrete finishing row, so this page uses it as context only and makes no finishing claim from it.

Heavy Industry & Fabrication holds five rows, all Hard and all Available on request: concrete work, glass manufacturing, metal fabrication, operating heavy machinery controls and welding. The environment’s listed examples are metal fabrication, woodworking, glass manufacturing, a concrete plant and a welding workshop. Those are examples for the whole environment, so they do not say where a concrete work batch is recorded. The catalog names no edging, jointing, brooming or curing, and no power trowel or bull float. Anything beyond the row’s own description is a skill-spec question for a Spec Pilot.

Difference

How is this different from the stereo, masonry and drywall pages?

The stereo concrete finishing page covers the same row through a different lens: what a calibrated stereo pair adds when the slab surface is the subject, and what MCAP is. This page covers the egocentric capture of the row itself: the first-person view, the streams, the labels and how to order. If your question is about depth from a stereo pair or about MCAP delivery, start with the stereo page. If it is about the first-person recording of the task, start here.

The masonry and plastering page is framed around a search phrase and notes that the catalog has no bricklaying, blockwork, mortar or plastering row, so it covers the concrete row as written, as the only row in that family. This page adds no masonry, bricklaying or plastering claim either. The drywall and wall finishing page covers a different task: hanging, taping and mudding drywall in the Active Construction Site environment, rated Medium. The comparison table below sets the pages side by side.

Why first-person

Why record concrete finishing from the first-person view?

Finishing is a sequence of passes with a hand tool over a surface. A shovel, a float and a trowel are each held differently, and the useful signal is how the hands and the tool move through the sequence, in what order, and how one step hands over to the next. A first-person camera keeps both hands, the tool in the hand and the surface in the same frame, from a viewpoint close to what a robot’s head or wrist camera would have.

Hand pose is delivered for both hands as 21 joints per hand in three dimensions, with a per-joint visibility flag for occluded fingers. A float or trowel is gripped so that fingers wrap the handle, and a shovel grip can be hidden by the load, so the flag marks the joints that were not visible instead of presenting a guess as a measurement. Contact events mark when a hand meets the tool or the item, and wrist cameras can be added in the spec for a grip the head camera loses.

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. What this data gives you is human first-person demonstrations in which the hands, the work and the step structure are explicit. This page does not claim that any robot performs concrete finishing, that any model improves from this data, or any result.

Episode contents

What does each episode contain?

Each attempt is delivered as one episode. 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. For the instrumented tier the site describes 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 catalog lists one capture setup for the concrete row, a head-mounted phone or a multiview rig, and which tier a program uses is set in the skill spec.

Every stream is hardware-triggered onto one PTP clock domain, so depth, motion and hand pose line up with the same video frame. Calibration is measured twice per session, and an episode is accepted only if RMS reprojection stays under 0.28 px. The only language field the site ties to Firsthand capture is one task instruction per episode. The site does not list a force or tactile stream for the concrete row, or any measurement of surface quality.

Labels and delivery

How is the data labeled and delivered?

Inside each episode, action segments divide the timeline into spans, each carrying start_ns, end_ns, verb and noun. For this task a segment can be one step as written in the skill spec, such as a pour, a screed pass, a float pass or a trowel pass, so one pour-and-finish job becomes a sequence of labeled steps rather than one long clip. The other standard fields are hand_joints and hand_visible for 3D hand pose, contact_events, objects, and per-frame instance masks where the spec calls for them, stored at the depth resolution of 848×480 so they line up with the depth stream.

Delivery is in the format your stack already reads: RLDS, LeRobot, WebDataset, HDF5, zarr or Rerun .rrd, or raw streams with metadata. The delivery formats table below gives the official description of the three that have an official source. In the HDF5 layout the site documents, hand pose sits at annotations/hand_joints and action segments at annotations/action_segments.

The site does not list labels for the quality of a finished surface. There is no stated label for whether a surface is level, flat, worked or finished to a grade. The verb and noun on each action segment is the vocabulary the skill spec defines with you, and anything beyond it is a scoping conversation, not a delivered field.

Stereo

Is this stereo capture?

Not as the catalog states it. The concrete row lists "Phone (head-mounted) / multiview rig" as its capture setup, and the site does not say a stereo pair is already running on it. Stereo applies through the site’s general stereo offer: a head-mounted stereo rig is a capture choice agreed in the skill spec and confirmed on a Spec Pilot. The stereo concrete finishing page covers that choice, the depth behaviour the site states for wet and reflective surfaces, and MCAP delivery.

The site defines two capture modes. A head-mounted phone records the first-person view of real work on site. A multiview rig with handheld grippers records a staged working set. The concrete row lists both and does not say which applies, so the site does not settle whether a given concrete batch is real work at a real site or a staged set. That is agreed in the spec. The site names no plant, contractor, site or venue for this row.

Statuses

What do "Available on request" and "Hard" mean for concrete?

Available on request means the work is scoped and recruited to your spec when you ask, rather than already queued. For the concrete row that means contributors with the right access are found after the spec is written. Ready to deploy, the status of the general trades row, means contributors and site access are set up and capture starts once a spec is signed. The concrete row is rated Hard in the catalog. The site gives no breakdown of what makes it Hard, and this page does not invent one.

Quality

How is quality checked on concrete episodes?

Validation runs against signed accept criteria written into the spec before capture. An episode is rejected, among other criteria, when the manipulated object is out of frame for more than 15 percent of the episode, when more than 35 percent of depth pixels in the manipulation region are invalid, or when an undocumented bystander appears. A human reviewer then checks each episode against the task spec.

A reject log ships with each batch, only footage that clears the criteria is billed as a validated hour, and rejected items are recaptured at the provider’s cost. This page states no pass rate or reject rate for concrete work, because the site states none.

Privacy

How is privacy handled on a concrete finishing capture?

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. At ingest, automated detection finds faces, plates and screens and blurs them irreversibly, and speech is scrubbed of names, addresses and card numbers. A human reviewer confirms that no un-consented person is identifiable in any frame, and footage that cannot be de-identified is discarded. An undocumented bystander is also a machine-checked reject criterion. The anonymization record is delivered with the data.

If your program needs terms about premises, access or who may be in frame, raise them when you scope it. The site states nothing about safety, compliance, certification or trade licensing for this row, and this page makes no such claim. None of this is legal advice, and the executed agreement controls. Capture at your own facility is listed as not offered.

Speed and scale

How fast can a custom concrete finishing dataset be collected?

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. The site does not say whether coverage for the concrete row is deep or thin, so that is settled in your quote. After the first batch, on-demand top-ups run against the same accept criteria, and the site does not state a turnaround for those later batches.

Scale comes from custom collection at scale through a network of thousands of capture providers, for a consistent, recurring data flow to labs. Collections are billed per validated hour. The site states no batch size, cadence, provider count or hours for concrete work, so those are Not stated and are set in your quote.

Ordering

How do you order an egocentric concrete finishing dataset?

Describe the steps and their order, what counts as a completed pass, the setting and the variation 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.

Three choices matter most. Name the step and the item rather than a broad activity, using catalog wording such as "mix, pour, screed and finish concrete into forms", because that is what lets each step become one action segment. Name the setting and the capture tier, since the row lists a head-mounted phone or a multiview rig and the choice is yours to settle. And write the reject rule and the acceptance test first, so the batch is judged against terms you agreed.

Capture and delivered fields

What is captured and delivered for a concrete finishing episode?

The capture setup and delivered fields as stated on this site, applied to the concrete row. Counts of episodes, sites, contributors and hours are set in the skill spec.
PartAs stated on this site
Data typeHuman first-person capture by paid, consented contributors; not robot or teleoperation data
Instrumented rigHead-mounted stereo rig worn by the contributor; first-person RGB
StreamsTen: 4K head camera, two wrist cameras, depth, IMU, 3D hand and body pose, instance segmentation, action labels
SyncHardware-triggered onto one PTP clock domain; aligned under 2 ms
DepthMetric depth recovered from the calibrated stereo rig
IMU6-DoF at 200 Hz from head- and wrist-mounted sensors
Hand pose21 joints per hand in 3D, both hands, with a per-joint visibility flag (hand_joints, hand_visible)
Contact eventsMarks when a hand meets the tool or the item (contact_events)
Action segmentsstart_ns, end_ns, verb and noun per step as written in the skill spec (action_segments)
Language fieldOne task instruction per episode
CalibrationTwice per session; accepted only under the 0.28 px RMS reprojection gate
Catalog capture setupPhone (head-mounted) / multiview rig, as listed for the concrete row

Every collection runs through the same seven-stage end-to-end custom collection pipeline.

Catalog rows

Which catalog rows name concrete?

The two rows that name concrete, with the catalog description, environment, skill, capture setup and status as listed. Only the first is a concrete finishing row.
Task (as listed)Description (as listed)EnvironmentSkillCapture setupStatus
Concrete workMix, pour, screed and finish concrete into forms using shovels, floats and trowels.Heavy Industry & FabricationHardPhone (head-mounted) / multiview rigAvailable on request
Construction General Skilled TradesSkilled trade work on active sites including framing, concrete and structural tasks.Active Construction SiteHardPhone (head-mounted) / multiview rigReady to deploy

Cluster map

How does this page differ from the stereo, masonry and drywall pages?

The nearest pages in the construction cluster and where this one differs. The stereo page covers the same row; the masonry and drywall pages are different framings or tasks.
PageFramingWork coveredConcrete coverage
Egocentric concrete finishing data (this page)First-person capture: head-mounted phone or multiview rig as the row lists, ten-stream stack in the instrumented tierMix, pour, screed and finish concrete into forms with shovels, floats and trowelsThe "Concrete work" row; the general trades row as context only
Stereo concrete finishing dataCalibrated stereo pair and MCAP delivery for the same rowThe same concrete rowThe "Concrete work" row and the general trades row
Egocentric masonry and plastering dataFramed around a search phrase; the catalog has no bricklaying, blockwork, mortar or plastering rowThe concrete row as written, the only row in that familyCovers the same row as context; adds no masonry claim
Egocentric drywall and wall finishing dataFirst-person capture of wall tasksHanging, taping and mudding drywall; Active Construction Site, rated MediumNone; a different task

Delivery formats

Which delivery formats are offered, and where are they documented?

The delivery formats the site states, with the official source read for this page where one is linked. Formats without a linked source are not described further here.
FormatAs the official source describes itOfficial source read
RLDSAn ecosystem of tools to store, retrieve and manipulate episodic data, including for learning from demonstrations and imitation learningRLDS repository (GitHub)
LeRobotDataset format of synchronized MP4 videos or images for vision and Parquet files for state and action dataLeRobotDataset v3.0 (Hugging Face docs)
WebDatasetTar files in which files that belong together share the same basename and make up a training sampleWebDataset repository (GitHub)
HDF5, zarr, Rerun .rrd, raw streams with metadataOffered on this site; not described further hereNot linked on this page

Skill spec

What does the skill spec fix for an egocentric concrete finishing order?

Spec element and where a buyer can see it in the delivery. "Set per program" means the site states no fixed value.
Spec elementWhat it decidesWhere it shows up
Finishing step and orderWhat one pass is: a pour, a screed pass, a float pass, a trowel passEpisode boundaries and action segments; set per program
Setting and capture modeReal work with a head-mounted phone, or a staged set with a multiview rigSet per program; the site does not settle it for this row
Rig and streamsHead-mounted view, wrist views, hand pose, IMU, and a stereo pair if you want oneThe channels you can read; confirm on the Spec Pilot
Delivery formatHow the data is packagedRLDS, LeRobot, WebDataset, HDF5, zarr, Rerun .rrd or raw streams
Accept criteriaWhat passes QAReject log per batch
PrivacyHow people in frame are protectedFaces and screens blurred at ingest
LicenseWho owns the datasetPerpetual, irrevocable, buyer-owned, non-exclusive by default

Not stated

What does the site not state for concrete finishing?

Each item below is something a buyer may ask about and the site does not state, so this page does not either.
ItemWhat the site states
Edging, jointing, brooming, curing, power trowel, bull floatNot stated; the catalog names none of them
Slab size, form type, mix, surface, tolerances or strengthsNot stated
Task duration, episode count, site count, hoursNot stated; set in the skill spec
Real work or staged set for this rowNot stated; the capture setup lists both modes
Plant, site, region or city for this rowNot stated
Flatness, level or finish-quality labelsNot stated; no such delivered field is listed
Force or tactile streamNot stated; not among the listed streams
Stereo already running on the concrete rowNot stated; a stereo pair is agreed in the skill spec
Safety, compliance, certification or trade licensingNot stated
Turnaround for the concrete row specificallyNot stated; the site states a median of 48 hours where coverage is deep and three to four weeks where it is thin
Price list, minimum order, batch sizeNot stated; set in your quote

Example footage

What does first-person capture look like?

One real clip already published on this site, shown as a format and rig example: ten seconds of first-person home cooking with faces blurred. It does not show concrete work, screeding, floating or troweling.

egocentric_cooking_10s.mp4 · first-person view · faces blurred · 1080 × 1080 web preview · format example, not a concrete clip

No concrete finishing clip is published on the site, so this is labelled as a format example. Request a sample to see delivered files. See sample data outputs, request an egocentric data sample.

FAQ

Egocentric concrete finishing data, answered.

What is egocentric concrete finishing data?
Custom first-person video of a person mixing, pouring, screeding and finishing concrete into forms with shovels, floats and trowels, captured by a paid, consented contributor working to a written skill spec, and delivered with synchronized streams, hand pose, contact events, action segments and consent and anonymization records.
Which catalog row covers concrete finishing?
"Concrete work" in the Heavy Industry & Fabrication environment: "Mix, pour, screed and finish concrete into forms using shovels, floats and trowels." It is rated Hard, lists "Phone (head-mounted) / multiview rig" as its capture setup, and is Available on request.
Is this stereo capture?
Not as the catalog states it. The row lists a head-mounted phone or a multiview rig. A stereo pair applies only through the site’s general stereo offer and is agreed in the skill spec. The stereo concrete finishing page covers that choice.
Do you provide labels for flatness, finish quality or cure state?
Not stated. The delivered fields are action segments, 3D hand pose with visibility flags, contact events, objects and instance masks where the spec calls for them. No flatness, level, finish-quality or cure label, and no force or tactile stream, is listed.
Is the data recorded at a real site or on a staged set?
Not stated for this row. The site defines a head-mounted phone as recording the first-person view of real work on site and a multiview rig with handheld grippers as recording a staged working set. The row lists both, and which applies is agreed in the spec.
How is privacy handled?
Every contributor signs a written release before capture. At ingest, faces, plates and screens are detected and irreversibly blurred, speech is scrubbed of names, addresses and card numbers, and a human reviewer confirms the result. Footage that cannot be de-identified is discarded, and the anonymization record is delivered with the data.
How fast is a first batch, and who owns the data?
Where coverage is deep, the median time from signed spec to first validated batch is a median of 48 hours; thin coverage runs three to four weeks. Which applies to the concrete row is not stated. The buyer owns the dataset under a perpetual, irrevocable, non-exclusive-by-default licence, with exclusivity available.

Sources

Where are the delivery formats documented?

Request a custom concrete finishing dataset.

Tell us the finishing steps and their order, what counts as a completed pass, the setting and the streams you want. We will quote reach, timeline and price against a written spec, with a first validated batch in a median of 48 hours where coverage is deep.