Egocentric Paper Trimming and Collating Data

Egocentric paper trimming and collating data is real-world video of a person working with sheets of paper at a desk-and-document station: lining paper up on a trimmer, cutting it, then gathering multi-page sets in order and stapling each one. The video is first-person, so it shows the hands, the sheets, the trimmer and the stapler as the person works. Alongside the video, each episode follows the Firsthand episode schema, with depth, 3D hand pose, contact events and frame-accurate action segments aligned on one hardware clock.

Teams search for it under several names, including paper trimming robot training data, egocentric guillotine trimmer video and a staple and collate dataset. They all describe first-person demonstrations of the same office jobs, collected to a written spec. This page covers only what the Firsthand site states about it, and marks everything else as Not stated.

Two rows in the site task catalog cover this work, both in the Office Document Station environment. The first is "Trim with guillotine", described as "Align and trim paper on a trimmer." The second is "Staple + collate sets", described as "Collate multi-page sets and staple each." Both are rated Easy, both use a multiview rig with handheld grippers, and both are listed as Available on request.

Anything outside those two rows is Not stated. That includes the type of trimmer, the type of stapler, the kind of paper and any other finishing step, so none is claimed here. If your policy needs one of them, you write it into the spec and we scope it, because the catalog rows are a starting vocabulary, not a fixed menu.

The site defines Available on request as work that is scoped and recruited to spec when a buyer asks for it. Neither row is listed as Active collection or Ready to deploy, so the site does not say that a trimming or collating capture stream is running today, and it does not state any delivered hours for them. A first batch is therefore quoted against your spec rather than drawn from a stream that already exists.

The site describes the Office Document Station as a desk-and-document station with a printer and tray, binders, a stapler, a label maker and mailers. Its rows use the multiview rig with handheld grippers, which the site describes 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 environment examples name a stapler but do not name a trimmer, even though the Trim with guillotine row sits in that environment. The site names no office, company or building, and does not state the number of sites, the regions or the venue type. It does not state capture at a customer own facility either. Environment, access and schedule are agreed in the written spec.

The catalog states the capture mode for both rows: a multiview rig with handheld grippers. That is the setup this page describes, and the site does not state a head-mounted camera for these two rows, so none is claimed. The site does state, in general, that a robot or humanoid perceives from a head- or wrist-mounted camera rather than a static overhead shot, and that egocentric capture is built to match that geometry. How that applies to a multiview rig is set in the spec.

The site does not state a per-environment stream list, so which of the ten streams an order carries is decided in the skill spec. Camera count, resolution, frame rate, desk layout, paper volume and the type of trimmer or stapler are Not stated, so raise any of them in the spec.

Each episode follows the Firsthand episode schema: ten streams synced under 2 ms, with head and wrist RGB, depth, 3D hand pose with per-joint visibility, contact and grasp events, and frame-accurate action segments. Squaring a stack against a trimmer, pulling a sheet from a pile and pressing a stapler closed are all hand and paper contacts, and the site pairs contact events on grasp and release with an IMU on one clock and with action segments.

Object labels use the terms your spec sets. Paper sizes or stock types, sheet counts, trimmer and stapler makes are Not stated. Task durations, cycle times and the number of episodes are Not stated too. Success or quality labels, such as whether a cut is straight or a set is in order, are Not stated, as are certifications and robot results.

The catalog rates both rows Easy, so the site does not call them Hard, and this page makes no claim about robot results. What the site does say is where paper work is difficult. Paper is thin and flexible, so a sheet or a stack changes shape every time it is touched and hides what is under it. Small items are easy to lose behind a fingertip, and collating repeats many near-identical motions that differ only in order.

Hand occlusion is the capture risk the site pairs with hands on small items, and it addresses it with 3D hand pose that carries per-joint visibility and with depth. The site gives no failure modes specific to a trimmer, so any it needs, such as a stack that is misaligned before the cut, are written into the spec as start states and accept criteria.

A good spec treats each named job as its own task. Aligning and trimming a stack, collating a set and stapling it each get a start state, an end state and a set of accept criteria. The site suggests variation in starting states, with a stack out of order as one example, and says a task prone to going wrong can carry failure and recovery cases that the spec requires on purpose.

Each step becomes a labeled action segment with a start and end timestamp, a verb and a noun, so a continuous run still yields individually labeled steps. The site notes that a spec can chain document station rows into one sequence if your policy needs the whole run, and that whether a chained sequence is collected as one episode or several is agreed in the spec.

Not stated. The site describes the Trim with guillotine row as "Align and trim paper on a trimmer" and says nothing about blade guards, cut safety rules, operator training or protective equipment for this work. It makes no safety-compliance claim for the document station rows. If your program has handling or safety requirements for a trimmer, state them in the spec before capture begins so they are part of the signed accept criteria, and expect them to be scoped as part of the quote.

Delivery formats are RLDS, LeRobot and WebDataset, plus HDF5, zarr and Rerun .rrd. These episodes use the same schema as every other Firsthand collection, so a loader that reads another Firsthand dataset reads this one without changes. The format table below sets each format against what its official documentation states.

Every batch is validated against signed accept criteria before it ships, with a reject log, and delivery is rolling, per validated batch. Rejected items are recaptured at our cost, and billing is per validated hour.

Every episode links to a signed contributor release through a documented consent chain, and the consent record is part of the delivery. A vision model runs on every frame at ingest, flagging faces, heads, licence plates and any legible screen or document. The site describes the documents class as phone and monitor content, forms, badges and mail that would leak names, addresses or account data. Those regions are irreversibly blurred, and a human reviewer confirms the result before footage ships.

If the reviewer is unsure a region is fully covered, the episode is held or discarded, never downgraded and shipped. So legible names, addresses and document content are not in the delivered pixels. If your pipeline needs document content, raise it before the spec is signed; whether any engagement can keep specific content is Not stated. The site states no other privacy, compliance or legal terms for this work, and no in-home capture.

The off-the-shelf inventory on this site lists no paper trimming or collating dataset, so this is custom collection. Custom collection at scale runs through a network of thousands of capture providers, for a consistent, recurring data flow to labs. The site states no provider count, hours or throughput for these two rows, so none is claimed here.

Where coverage is already deep, the first validated batch lands in a median of 48 hours from signed spec. Thin environments typically take three to four weeks, because vetting operators there is the bottleneck, and we quote that up front. The site does not state how deep coverage is for these two rows, and they are Available on request, so plan on a quote rather than assuming the 48 hour figure.

The site also states program milestones. A pilot reaches its first delivery in two to three weeks, and a standing program reaches full rate in four to six weeks. Those assume capture coverage is already in place. On-demand top-ups run against the same spec and accept criteria after each training run. A batch size and a later-batch turnaround are Not stated, so both are quoted against your spec.

You describe the trimming and collating jobs, the environment and the sequence your policy needs. We turn that into a written spec covering tasks, environments, capture tier, annotation depth and accept criteria, then quote reach, timeline and price before any capture begins. This is the custom egocentric data collection route that every Firsthand task family follows.

A Spec Pilot, a small and fast first batch, can prove the spec and accept criteria before you scale. Every collection ships under a perpetual, buyer-owned commercial license, with a reject log and signed accept criteria on every batch.