Egocentric Shoelace and Eyelet Threading Data

It is first-person video of one person threading a lace through eyelets, recorded from that person’s own point of view while they work. A paid, consented contributor wears the capture rig and performs the 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.

The target query is a shoelace query, but the catalog row is written more narrowly. It says "Thread a lace through eyelets." It does not name a shoe, a boot, a lace type or a bow. This page keeps to the row as written and treats anything more specific, such as a particular lacing pattern or tying a finished bow, as a spec question settled in a Spec Pilot.

The page is about buying a custom collection of those demonstrations: how the row is specified, how the capture is set up, how each attempt is labeled and what the site does not state. It makes no claim about a trained model, a benchmark or a success rate.

One row counts. "Thread lace eyelets" sits in the Deformables & Textiles Bench, in the Manufacturing & Workshops sector, whose examples read "Cloth-and-cable bench: garments, towels, cables, form boards, bags." The row is described as "Thread a lace through eyelets." and is rated Hard. Its capture setup is "Multiview rig + handheld grippers" and its status is Available on request.

Available on request is the site’s wording for a task that is scoped and recruited to spec when a buyer asks for it, as opposed to Active collection or Ready to deploy. So this page does not claim delivered hours, a stock of footage or a published dataset for this task. It describes what can be collected for you.

The first table below reproduces the row as the catalog states it, with the neighbouring fastening rows from the same bench shown as context only. The environment page for the Deformables & Textiles Bench lists the same rows, and the catalog row is the authority if the two ever read differently.

No row in the catalog describes tying a shoelace or tying a knot. The existing rope and cord page says so plainly, and this page agrees: the lace row covers threading a lace through eyelets, and tying a bow after threading is not part of its description.

The row also names no footwear. The environment examples are garments, towels, cables, form boards and bags, and the row does not say whether the eyelets are on a shoe, a garment, a bag or a form board. Where the lace is threaded, how many eyelets there are and which lacing pattern applies are not stated, so a spec names them.

That matters for a buyer. If your policy needs a lace threaded through a shoe and then tied, the spec has to say so, with where each step starts and ends, and a Spec Pilot proves it against your conditions before you scale. If your policy only needs the threading pass, the row as written already describes it. The second table lists what is and is not stated.

Three further rows on the same bench are fastening or cable tasks and are mentioned here as context only, not as claims of this page. "Fasten buttons/snaps" is described as "Fasten a row of buttons or snaps." and is rated Hard. "Open/close zipper" is described as "Start and run a zipper closed." and is rated Medium. "Coil + Velcro cable" is described as "Over-under coil and strap a cable." and is rated Medium. A fourth, "Route + clip cables", routes cables to a form board and is rated Medium.

All of them use the same capture setup, "Multiview rig + handheld grippers", and are Available on request. The bench lists twelve rows in all; the rest cover folding, flattening, hanging, socks, a duvet cover and wrapping, which handle fabric but do not thread anything.

If your project needs several of these fastening tasks together, the buttons, zippers and fasteners page treats the four fastening rows as its subject. This page takes the lace row on its own, so the spec language, the labels and the failure modes below are written around threading.

The site makes the central point in its own words: a cable has no single shape, and a lace shares that property. A lace is a long, thin, flexible object, so its form changes as the hands move it, and a camera sees a different shape every few frames. Three consequences show up in the data, and each is something the site already addresses.

First, occlusion. Fingers disappear behind objects in manipulation, which is why 3D hand pose carries a per-joint visibility flag. In a threading task the fingers close around the lace end and the eyelet at the same time, so a hidden joint is the normal case rather than an exception. The flag marks joints that were not seen instead of presenting a guess as a measurement.

Second, coordination. One hand holds the item with the eyelets while the other feeds the lace, so both hands need to be in frame at once. First-person capture keeps both hands and the work in view across the whole attempt, and hand pose is delivered for both hands.

Third, tracking. 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. A lace is not named in that statement, so whether a lace task asks for them is a spec decision. Metric depth on sub-centimetre parts is flagged rather than interpolated, but whether an eyelet or a lace tip counts as a sub-centimetre part is not stated, and the site gives no mask accuracy figure for laces.

Firsthand captures people, not robots. There is no robot joint-angle log and no teleoperation data for this task, and using the episodes for a specific robot means your own retargeting step.

The catalog lists the capture setup per row, and for "Thread lace eyelets" it is "Multiview rig + handheld grippers". The site explains that setup as a staged working set recorded by a multiview rig while the operator performs the task with handheld grippers, so demonstrations map onto a robot end effector. The catalog lists no stereo pair for this row, so this page does not describe the capture as stereo.

The wider stack the site documents is ten streams: a head-mounted global-shutter 4K camera at 60 fps with a 120 degree lens, two wrist-mounted cameras at 1080p and 60 fps, depth, 6-DoF IMU data at 200 Hz, 3D hand and body pose, instance segmentation and action labels, hardware-triggered onto a single PTP clock domain and aligned under 2 ms. The rig is calibrated at the start and end of every session, and an episode is accepted only if RMS reprojection stays under 0.28 px.

Which of those streams apply to a given lace collection, how many cameras are used and what the grippers are is set in the skill spec. The site states no camera count, resolution or frame rate specific to this row, and no gripper detail beyond the sentence above. Force and tactile readings are not among the standard streams.

Each attempt is delivered as one episode, and a lace task needs a decision up front about what one attempt is: one eyelet, one pair of eyelets, one full lace threaded through every eyelet, or a threading pass followed by a finishing step. The site does not set that length, so the skill spec does.

Inside the episode, action segments divide the recording into start- and end-timestamped spans, each carrying start_ns, end_ns, verb and noun. The site states a taxonomy of 97 verbs and 512 nouns, annotated to the frame and human-reviewed. A spec that says only "thread the lace" gives you one long segment. A spec that names the individual steps gives you a sequence you can slice. As an illustration of how a spec can cut an episode, and not as steps the catalog states, it could name laying out the lace, passing it through an eyelet, pulling it through and moving to the next eyelet.

The other standard fields are hand_joints and hand_visible for 3D hand pose, 21 joints per hand with a per-joint visibility flag, contact_events for the moment a hand meets an item, objects, and instance_masks stored at the depth resolution of 848×480 so they line up with the depth stream. Every episode carries int64 nanosecond timestamps per stream, camera intrinsics and extrinsics.

The site lists no label for the lace itself: no threaded or unthreaded state, no eyelet count, no lace tension and no completion score, and no contact force reading. If your pipeline needs one, it is scoped in the spec rather than assumed.

Several existing pages touch this task. The buttons, zippers and fasteners page counts four fastening rows, and the lace row is one of them; this page takes that one row as its main subject. The rope and cord page covers the cord-like rows together and states that no knot or shoelace row is listed. The deformable object page covers laundry, bed-making and bag tie. The shoe and footwear repair page covers a different row, "Shoe repair", in the Repair Shop environment, and is about repairing a shoe rather than threading a lace.

The comparison table below sets those pages next to this one. It states each row as the site states it and makes no claim beyond that. If your question is about a single fastener type, this page and the buttons page overlap; if it is about cord handling in general, start with the rope and cord page.

Only what the site already states applies. Every participant signs a written release granting commercial training rights before capture begins, and consent artefact IDs link each episode to it. Faces, plates and screens are detected and irreversibly blurred at ingest, speech is scrubbed of names, addresses and card numbers while ambient audio is kept, and a human reviewer confirms that no un-consented person is identifiable in any frame.

Footage that cannot be de-identified is discarded rather than shipped, and the anonymization record is delivered with the data. A second person in frame follows the same rule, and the spec settles the details for your bench. This is not legal advice, and the executed agreement controls.

The site states one published timeline: from a signed spec to a first validated batch, a median of 48 hours where coverage is deep, and typically three to four weeks where it is thin and contributors need to be recruited first. The site does not say which timeline applies to the Deformables & Textiles Bench, and the row is Available on request, so ask when you request the quote.

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. Collections are billed per validated hour, rejected episodes and recaptures are never billed, and the site states no price list or minimum order for this task. Capture at your own facility is listed as not offered.

Start with the free, ungated sample pack to check the loader and the label quality, then move to a Spec Pilot: a small first batch against the written spec that proves the fill rate and the reject rate on your conditions before you scale.

A good brief names the task, the item the eyelets are on, the order of steps, how each step starts and ends, what counts as threaded, whether tying a bow is in scope, and what counts as a reject. The site’s skill-spec guide says to name the verb and the noun rather than a broad activity. Then request a custom dataset, and we quote reach, timeline and price against the spec.