The hands and the garment in one frame
A head-mounted view keeps both hands and the work in frame, close to what a robot’s onboard camera would see at a fitting. Hand pose, contact events and per-frame instance masks are recorded on one clock.
Now recruiting capture teams worldwide. Get paid to record first-person video of real work.
Apply to captureHard task
Why first-person here
A head-mounted view keeps both hands and the work in frame, close to what a robot’s onboard camera would see at a fitting. Hand pose, contact events and per-frame instance masks are recorded on one clock.
The site states that fingers disappear behind objects and that the hands occlude the work. 3D hand pose carries a per-joint visibility flag, so a finger hidden under a fold of cloth is marked as hidden rather than silently filled in.
The site says masks for deformable objects are shipped per frame rather than propagated from one frame to the next. A garment on a person is a deformable object, so the instance masks follow it frame by frame.
The row pins and marks the garment on the client, which puts a second person in the frame. Who is being fitted, and how they are consented, is settled in the spec, under the same rules as any other capture.
Definition
It is first-person video of a person doing the work named in one catalog row, pinning and marking a garment on the client and adjusting seams and length for fit, recorded from their own point of view with synchronized sensor streams alongside. People reach it through several phrasings: egocentric garment fitting data, first-person pinning and seam adjustment video, a tailoring fit robot dataset, or custom textile data collection. They describe one request from different angles.
The data is human capture. A paid, consented contributor wears the rig and performs the task to a written skill spec. It is not robot or teleoperation data, so using it for a specific robot still needs your own retargeting step, and it is not a fixed corpus you download: none of the site’s off-the-shelf dataset descriptions names garment fitting, so this is custom collection.
Catalog row
One row, in the Textile Arts & Stitching Studio environment of the Creative Workshops & Crafts sector, whose examples line reads "Hand embroidery, tailoring studio, quilting, needlepoint, custom garment fitting, hemming". It is named "Custom garment fitting" and reads "Pin and mark a garment on the client and adjust seams and length for fit." Its skill is Hard, its capture setup is Phone (head-mounted) / multiview rig and its status is Ready to deploy. The first table below reproduces it.
The site defines Ready to deploy as contributors and site access being set up, with capture starting once a spec is signed. It does not say that recording is running now, and this page does not call the row active or ongoing. It also claims no volume, hours or delivered clips for this row. The row names two actions and one setting. It names no particular fabric, tool or step, so anything more specific goes into the skill spec.
Hard row
On this site, Hard is the skill level shown in the catalog’s Skill column, one of Easy, Medium or Hard, and this page uses the word in no other sense. The row carries it. The hard-task hub lists every row rated Hard, grouped by environment, and this row is one of them. In the same studio, the Tailoring row is also rated Hard, while hand embroidery, hemming, needlepoint and quilting are rated Medium.
What the row text itself shows is a description of two connected actions, marking the garment and adjusting it, carried out on a client. The site publishes no failure rate, tolerance or claim about workers’ skill for this row, so this page makes none.
Difference
The stereo embroidery and tailoring page is broader. It covers all nine Textile Arts & Stitching Studio rows as a stereo and MCAP question, so custom garment fitting is one of nine rows there. The egocentric embroidery and hand stitching page covers the hand-stitching slice of the cluster as first-person video. The egocentric sewing and textile handling page covers cutting, pinning, hemming, stitching and handling fabric across the catalog. None of them has the fitting row as its main topic, which is why the hard-task hub now points that row here.
The site has no separate egocentric tailoring and alterations page, so none is compared here. This page is about the one fitting row only, in first-person framing. The comparison table below sets the pages side by side.
Capture setup
The catalog lists "Phone (head-mounted) / multiview rig" for the row. The catalog describes the head-mounted phone as recording the first-person view of real work, and a multiview rig adds synchronized third-person angles when the spec asks for them. The catalog does not say which of its two modes applies to a garment fitting program, so that is confirmed in the skill spec.
The site also describes a ten-stream instrumented tier: a head-mounted stereo rig worn by the contributor, with first-person RGB and metric depth recovered from the calibrated rig. Whether a garment fitting program uses it is Not stated. The site states no stereo pair for the studio rows, so stereo is a capture choice agreed in the spec and confirmed on a Spec Pilot. The site lists capture at a customer’s own facility as not offered, so the setting is agreed in the spec. The capture setup table below lists the figures the site states.
Episodes
Each attempt is delivered as one episode. On this site the standard episode carries action segments, 3D hand pose for both hands with a per-joint visibility flag, contact events that mark when a hand meets an item, an objects list, and per-frame instance masks stored at the depth resolution of 848×480.
Force and tactile readings are not among the listed streams, and neither is any label for whether a fit is acceptable. If your model needs one of those, say so in the spec before capture, because the site does not promise it.
Why it is demanding
The reasons come from the row text and existing site copy only. First, the row is rated Hard in the catalog. Second, the row describes a garment being worked on a person, so cloth, hands and a second person share the frame. Third, the site states that the hands occlude the work and that fingers disappear behind objects. That is why hand pose carries a per-joint visibility flag, why contact events mark when a hand meets an item, and why masks for deformable objects are shipped per frame.
The site states that depth degrades on wet and reflective surfaces and that metric depth on sub-centimetre parts is flagged rather than interpolated. Whether anything in a fitting counts as such a surface or part is Not stated, and so is a mask accuracy figure.
Client and consent
The catalog row pins and marks the garment on the client, which puts a second person in the frame. The same rule applies to that person as to any capture: an un-consented person who is identifiable in a frame means the footage is de-identified or discarded, so settle who is being fitted, and how they are consented, in the spec.
Every contributor signs a written release, and each delivered episode links to it through consent artefact IDs in the batch data card. The consent chain has six checkpoints, and a break at any one discards the footage. At ingest, automated detection finds faces, plates and screens and blurs them irreversibly. Speech is scrubbed of names, addresses and card numbers, while ambient audio is kept. A reviewer confirms that no un-consented person is identifiable in any frame, and footage that cannot be de-identified is discarded. The anonymization record ships with the data. The site states nothing further about the client, and this page adds nothing to it. This is not legal advice, and the executed agreement controls.
Steps
The row gives two actions, not a list of steps, so the steps come from the spec. A spec that says "fit the garment" gives one long action segment. A spec that names each part you want separated, such as the pinning and marking, and the seam and length adjustments, gives a sequence you can slice by step, each with start_ns, end_ns, verb and noun.
Whether an episode covers one action or a longer run is a spec decision, not a default. The site sets no episode length for this row. The catalog does not name any sub-step, so this page names none and gives no fitting procedure. What counts as a finished fitting is Not stated and is written into the spec as accept criteria before capture.
Labels and QA
Validation runs against signed accept criteria written into the spec before capture, and a reject log ships with each batch. Episodes are machine-checked against stated reject criteria, including a sync breach, calibration drift, the manipulated object out of frame for more than 15 percent of the episode, an undocumented bystander, and a mismatch against the named skill spec. A human reviewer then checks each episode.
Only footage that clears the criteria is billed as a validated hour, and rejected items are recaptured at our cost. For this row, the spec decides what the manipulated object is: the garment, a part of it or a marking.
Not stated
Fabric types, tool names and brands, garment types, measurements, cycle times, error rates, and what a correct fit means are Not stated. They are set in the written spec, and this page claims none of them. The site also states no result from training any model on this data, and names no tailor, client, studio or customer for this row. It states no contributor safety protocol and no trade certification for any sewing row, and this page describes none.
Scale and speed
The October 2026 off-the-shelf catalog lists ~324,800 hours across 18 datasets. None of the site’s off-the-shelf dataset descriptions names garment fitting, so this is custom collection, not an off-the-shelf purchase.
It is custom collection at scale through a network of thousands of capture providers, for a consistent, recurring data flow to labs. 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. Which applies to this row is confirmed when the spec is quoted rather than assumed. At program level, a pilot runs two to three weeks to first delivery and a standing program reaches full rate in four to six weeks. Collections are billed per validated hour, and the site publishes no price list, minimum order or batch size for this row, so those are set in your quote.
Delivery
Delivery is in the format your stack already reads: RLDS, LeRobot, WebDataset, HDF5, zarr or Rerun .rrd, or raw streams with metadata, with readable converter source alongside so the schema can be retargeted to your own loader.
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.
Ordering
Describe where each fitting starts and ends, what counts as a finished step, and whether an episode covers one action or a longer run. Add the setting and lighting, who is being fitted and how they are consented, the privacy rule for the studio, and the variation your pipeline is missing. Then name the streams you want and the delivery format.
Most programs begin with a Spec Pilot, a small, fast 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. For the full list of rows rated Hard, see the hard-task hub linked below.
At a glance
Every collection runs through the same seven-stage end-to-end custom collection pipeline. End-to-end custom data collection is a managed service that takes an AI data need from problem to owned dataset in a single accountable pipeline.
Catalog row
| Task (as listed) | Description (as listed) | Environment | Skill | Capture setup | Status |
|---|---|---|---|---|---|
| Custom garment fitting | Pin and mark a garment on the client and adjust seams and length for fit. | Textile Arts & Stitching Studio | Hard | Phone (head-mounted) / multiview rig | Ready to deploy |
Compared pages
| Page | Main topic | Capture framing | Treatment of the Custom garment fitting row |
|---|---|---|---|
| This page: egocentric custom garment fitting data | The Custom garment fitting row only | First-person, head-mounted phone or multiview rig as the catalog lists it | The whole page: scope, limits, labels, client and ordering |
| Stereo embroidery and tailoring data | All nine Textile Arts & Stitching Studio rows | Stereo pair, agreed in the skill spec; not stated on the rows | One of nine rows, treated as a stereo and MCAP question |
| Egocentric embroidery and hand stitching data | Hand-stitching rows, including Hand embroidery, Needlepoint, Embroidery, Hemming and Tailoring | First-person | Not covered as a fitting page |
| Egocentric sewing and textile handling video data | Cutting, fitting, hemming, stitching and fastening fabric across the catalog | First-person | One row among several; the client rule is noted |
| Hard task data collection | Every catalog row rated Hard, grouped by environment | As each row lists it | Row listed in the hub table with its status and capture setup |
Capture setup
| Part | As stated on this site | For this row |
|---|---|---|
| Catalog capture mode | Head-mounted phone for the first-person view of real work; a multiview rig adds synchronized third-person angles when the spec asks for them | Stated in the catalog; which mode applies is confirmed in the skill spec |
| Instrumented tier | Head-mounted stereo rig worn by the contributor; first-person RGB with metric depth recovered from the calibrated rig | Whether a program uses it: Not stated |
| Streams (instrumented tier) | Ten: 4K head camera, two wrist cameras, depth, IMU, 3D hand and body pose, instance segmentation, action labels | Stated; set in the spec |
| Sync | Hardware-triggered onto one PTP clock domain; aligned under 2 ms | Stated |
| Hand pose | 21 joints per hand, 3D, in metres in the head-camera frame; per-joint visibility flag | Stated |
| Calibration | Twice per session; accepted only under the 0.28 px RMS reprojection gate | Stated |
| Masks | Per-frame instance masks, stored at 848×480; masks for deformable objects are shipped per frame | Stated |
| Setting | Capture at a customer’s own facility is not offered | Setting agreed in the skill spec |
Demands and streams
| Challenge | What the site states | Stated stream or label |
|---|---|---|
| Fingers hidden by cloth | The hands occlude the work; fingers disappear behind objects | 3D hand pose with per-joint visibility flags; per-frame instance masks; reject if the object is out of frame for more than 15% of an episode |
| Cloth that changes shape | Masks for deformable objects are shipped per frame rather than propagated | Per-frame instance masks |
| A second person in frame | An un-consented person who is identifiable in a frame means the footage is de-identified or discarded | Consent artefact IDs per episode; anonymization record; undocumented bystander is a stated reject criterion |
| Hand meets garment | Hand and work are in close contact | Contact events mark when a hand meets an item; no finer contact label is stated |
| Fit quality | No such label is listed | Not stated: completion is defined in the spec and episode boundaries |
| Force and contact pressure | No force or tactile stream is listed | Not stated |
Delivered vs not stated
| Item | Status | Detail |
|---|---|---|
| Action segments | Delivered | start_ns, end_ns, verb, noun, one per step |
| 3D hand pose | Delivered | 21 joints per hand; per-joint visibility flag for occluded fingers |
| Contact events | Delivered | The moment a hand meets an item |
| Per-frame instance masks | Delivered | Stored at 848×480 |
| Consent and anonymization records | Delivered | Consent artefact IDs per episode; anonymization record shipped with the data |
| Fabrics, tools, garment types | Not stated | Agreed in the skill spec |
| Measurements, cycle times and throughput | Not stated | No such figure is listed |
| Number of sites, episodes and hours | Not stated | Set per program in the skill spec |
| Fit or success labels | Not stated | No such label is listed |
Delivery and licensing
| Item | As stated on this site | Note |
|---|---|---|
| Formats | RLDS, LeRobot, WebDataset, HDF5, zarr, Rerun .rrd, or raw streams with metadata | Chosen in the skill spec; converter source ships with the data |
| License | Perpetual, irrevocable, worldwide, buyer-owned, non-exclusive by default | Exclusivity is available as a separate term; the executed agreement controls |
| Billing | Per validated hour | Rejected items are recaptured at our cost |
| Privacy | Faces, plates and screens blurred irreversibly; footage that cannot be de-identified is discarded | Anonymization record delivered with the data |
| Timeline | a median of 48 hours to first validated batch where coverage is deep; three to four weeks where thin | Which applies to this row is Not stated; quoted against your spec |
| Price, minimum order, batch size | Not published for this row | Set in your quote |
Skill spec
| Spec element | What it decides | Where it shows up |
|---|---|---|
| Actions in scope | Whether pinning and marking, adjusting seams and length, or both are in scope | What the cameras see; set per program, no per-action label is stated |
| Tasks and order | What one step is and how steps are named | Episode boundaries and action segments; set per program |
| Episode boundary | One action or a longer run | Set per program; the site does not state an episode length for this row |
| Object definition | Whether the garment, a part of it or a marking counts as the object | Contact events and the out-of-frame reject rule; set per program |
| Capture mode | Head-mounted phone, multiview angles, wrist views and any others | The streams you can read; confirm on the Spec Pilot |
| Client and privacy | Who is being fitted, how they are consented and what may be in frame | Settled before capture; bystander handling is a stated reject criterion |
| Accept criteria | What passes QA, including what counts as a reject | Reject log per batch |
| Delivery format | How the data is packaged | RLDS, LeRobot, WebDataset, HDF5, zarr, Rerun .rrd or raw streams |
| License | Who owns the dataset | Perpetual, irrevocable, buyer-owned, non-exclusive by default |
Example footage
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 garment fitting or any textile work.
No garment fitting clip is published on the site, so this is the nearest first-person clip, labelled as a format example. Request a sample to see delivered files. See sample data outputs, request an egocentric data sample.
FAQ
Sources
Tell us where each fitting starts and ends, whether an episode covers one action or a longer run, the setting, who is being fitted and how they are consented, the privacy rule for the studio, and the streams and format you want. We will quote reach, timeline and price against a written spec.