Egocentric Sample Tube Capping and Centrifuge Loading Data

It is first-person video of one person doing tube work at a bench, 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.

People reach this topic through several phrasings: egocentric lab tube handling data, first-person centrifuge loading video, a sample tube capping and racking dataset, or custom wet-lab data collection. They describe one request from different angles. The work is the same: write the task down, collect it on demand, and receive validated episodes.

Four, listed in lib/commercial-catalog-data.ts under the Wet Lab & Clinical Bench environment in the Healthcare & Life Sciences sector, whose examples line reads "Burette, pipettes, balance, labware, phantom for any invasive step." "Cap, vortex + rack tubes" reads "Cap sample tubes, vortex-mix, and rack them in order." "Load + balance centrifuge" reads "Load tubes balanced in a centrifuge." "Fill + cap sample tubes" reads "Fill tubes to a mark and cap." Those three are rated Medium. "Rack + label tubes" reads "Label and rack a set of tubes" and is rated Easy.

All four list "Multiview rig + handheld grippers" as the capture setup, and all four are listed as Available on request. The site defines that status as work that is scoped and recruited to spec when a buyer asks for it. That is a statement about the catalog row, not a promise of a delivery date. The site states no delivered hours for these rows, so this page claims none. The first table below reproduces the four rows with the catalog wording.

This page is about the tube rows as its main subject. The egocentric lab pipetting and sample handling page covers fourteen catalog rows, five from Pharmacy & Laboratory and nine from Wet Lab & Clinical Bench, with pipetting first and the tube rows alongside. The stereo lab pipetting and sample handling page covers seven rows, including the same four tube rows, and treats them as stereo capture delivered as MCAP. The egocentric pharmacy and lab bench page spans medication dispensing, labeling and weighing as well as lab work.

The egocentric slide prep and plate streaking page covers two Hard microbiology rows, "Prepare a slide" and "Swab + streak plate". The egocentric container opening page covers opening jars, bottles and tubes, which is the opposite direction from capping. The second table below sets the pages side by side with each status as the site lists it. This page does not repeat pipetting, weighing, titration or slide rows.

The catalog lists "Multiview rig + handheld grippers" for all four rows. The site describes 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 site does not explain the grippers further, and it does not state a stereo pair for these rows. This page does not describe the tube rows as stereo capture, and the site’s stereo offer is a separate page.

The site states that every custom capture runs ten streams synced under 2 ms, with head and wrist RGB, depth, IMU, 3D hand and body pose, instance segmentation and action labels as the spec asks. It lists capture at your own facility as not offered, so the setting is agreed in the spec and arranged by Firsthand. Camera count, resolution and frame rate for tube work are Not stated.

Two rows name capping. "Fill + cap sample tubes" is described as filling tubes to a mark and capping, so it joins a filling step to a closing step. "Cap, vortex + rack tubes" is described as capping sample tubes, vortex-mixing and racking them in order, so capping is the first of three steps. Both are rated Medium. The catalog does not say how they differ in length or in what an episode contains.

The closing step is a small, two-handed movement: one hand steadies the tube and the other works the cap. The site does not state a cap type, closure style, tube type or tube size, and no volume or fill level beyond the words "to a mark". A spec that says "cap the tubes" gives one long segment. A spec that names filling, capping, vortex-mixing and racking as separate steps gives a sequence you can slice. Whether capping and vortexing are one segment or two is a spec decision.

Racking appears in two rows. "Cap, vortex + rack tubes" racks the tubes in order after capping. "Rack + label tubes" is the Easy row, described as labeling and racking a set of tubes. The word "set" means the unit of work is several tubes, and the catalog does not say how many. Item counts are Not stated, so the spec names the number of tubes in an episode if it matters.

Racking in order and labeling add a requirement the single-tube rows lack: the sequence has to be readable. An accept criterion such as "in order" needs a definition in the spec, for example what the order is based on. The site lists no rack type and no labeling method. It also states that tube labels, sample labels and container markings are not listed as blur categories, so if label text in frame matters, raise it in the spec so the handling is agreed before capture.

One row names it. "Load + balance centrifuge" is rated Medium and described as loading tubes balanced in a centrifuge. It is a placement task: tubes go into positions, and the arrangement has to balance. The catalog lists no centrifuge, rotor or adapter type, no speed and no run, and the row stops at loading. This page does not describe operating a centrifuge.

What counts as balanced is not stated on the site. For a first-person episode the spec should say what a finished load looks like, where the episode starts and ends, and whether unloading is in scope. The row says load, so unloading is not assumed. The hands and the tubes are small and sit inside an enclosure at the moment of placement, so the spec should also say which view is needed there and whether a multiview angle or wrist view is wanted.

The site states that the hands occlude the work and that fingers disappear behind objects. That is why each hand joint carries a visibility flag, why contact events mark when the hand meets an item, and why instance masks are recorded per frame, stored at the depth resolution of 848×480 so they line up with the depth stream. The site does not say whether a cap or a tube counts as the item for a contact event, so the spec settles it.

Two machine-checked rules matter at a bench. An episode is rejected if the manipulated object is out of frame for more than 15 percent of it, and if more than 35 percent of depth pixels in the manipulation region are invalid. The site states that depth degrades on wet and reflective surfaces, and that a confidence map is delivered so low-confidence regions can be masked. It does not say how transparent tubes, glass or liquid in a vessel are handled. If your policy depends on those, raise it at the start and settle it in a Spec Pilot.

Each attempt is delivered as one episode. The site does not set an episode length, so the spec decides whether one episode is one tube, one rack or a full run from filling to loading. Inside the episode, the synchronized streams run from start to end, and action segments divide it into labeled spans, each carrying start_ns, end_ns, verb and noun. The other standard fields are hand_joints and hand_visible for 3D hand pose, contact_events, objects and instance_masks.

The site lists no label for the contents of a tube, a liquid level, a sample identity, a balance result or a pass-or-fail score, and no contact force or tactile reading. It states 6-DoF IMU data at 200 Hz from head- and wrist-mounted sensors, and nothing more about motion. Nothing here reports a robot result, a benchmark or a model improvement.

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. The consent chain has six checkpoints, and a break at any one discards the footage rather than shipping it at a lower grade. At ingest, automated detection finds faces, plates and screens and blurs them irreversibly. 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 reject criterion, and the anonymization record is delivered with the data.

The environment line lists a phantom for any invasive step, and this page covers no real specimens and no patient data. It gives no biosafety, contamination-control, clinical or laboratory-procedure instructions, and it makes no regulatory, sterility or certification claim. The site states no safety protocol for these rows. None of this is legal advice, and the executed agreement controls.

Tube, cap, rack and centrifuge types or brands, rotor and speed values, volumes, item counts, task times, error rates, the number of sites and what a correct or finished result means are Not stated. So are the camera count, resolution and frame rate for these rows, a lighting setup and the way grippers are held. They are set in the written spec, and this page claims none of them.

The catalog lists no uncapping row, no centrifuge unloading row and no well-plate or microplate row. If you need one of those, say so in the request and settle it in a Spec Pilot rather than assuming it.

Delivery is in the format your stack already reads: RLDS, LeRobot, WebDataset, HDF5, zarr or Rerun .rrd, or raw streams with metadata. Readable converter source ships alongside the data so the schema can be retargeted to your own loader. The official documentation for several of these formats is linked in the sources below. The license is perpetual, irrevocable and worldwide, buyer-owned and non-exclusive by default, with exclusivity available as a separate term.

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 which applies to the Wet Lab & Clinical Bench environment, so the timeline is quoted against your spec. A pilot reaches its first delivery in two to three weeks, and a standing program reaches full rate in four to six weeks. Speed comes from a warm pipeline, not from cutting consent, review or anonymization.

The October 2026 off-the-shelf catalog lists ~324,800 hours across 18 datasets. The site names no tube handling dataset among them, so this work 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. Collections are billed per validated hour, and rejected episodes and recaptures are never billed.

Describe the steps and where each starts and ends: filling, capping, vortex-mixing, labeling, racking and loading. Say whether an episode covers one tube, one rack or a full run, what counts as capped, in order, labeled or balanced, whether unloading is in scope, the setting, and the privacy rule for the bench and for anyone in frame. Name the variation your pipeline is missing. Then name the delivery format your stack reads.

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 fifth table below shows what each spec element decides.