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

Apply to capture

Task family

Egocentric Rack Rail and Cage Nut Mounting Data

Fastening rail brackets with a driver, and clipping cage nuts and mounting a blanking panel, are two rows in the site’s task catalog, both in the Data Center & IT Hardware Bench environment. This page covers egocentric server rack installation data for those two rows only: what the catalog says about them, how they are captured and listed, what a mounting episode can carry, and what the site does not state. We run custom data center data collection fast and on demand against a written spec, as custom collection at scale through a network of thousands of capture providers, for a consistent, recurring data flow to labs.
Catalog rows
Two: Screw-mount rack rails (Easy) and Clip cage nuts + panel (Medium)
Status
Both Available on request, as the catalog lists it
Catalog capture setup
Multiview rig + handheld grippers
License
Perpetual, buyer-owned

Requirements

What does rack rail and cage nut mounting data need?

01

Two named mounting tasks

Fastening rail brackets with a driver and clipping cage nuts before mounting a blanking panel are different sequences, so a spec names each one, where it starts and where it ends.

02

The capture mode set per task

The catalog lists a multiview rig with handheld grippers for both rows, so the spec states which streams that mode carries rather than assuming them.

03

Step boundaries on the same clock

Where the spec asks for them, contact events and action segments mark when a bracket, a cage nut or a panel meets the rack and where each step ends, aligned to video and depth.

04

Hidden parts recorded as hidden

Where hand pose is part of the spec, a per-joint visibility flag records which joints were hidden behind a bracket, a panel or the driver.

05

Rights you can ship on

A perpetual, buyer-owned commercial license backed by a documented consent chain.

Definition

What is egocentric rack rail and cage nut mounting data, and why does it need its own page?

It is video of the mounting work at a server rack: fastening rail brackets with a driver, and clipping cage nuts and mounting a blanking panel. People reach it through several phrasings: egocentric server rack installation data, first-person rack rail and cage nut video, a blanking panel mounting robot dataset, or custom data center data collection. It is recorded against a written spec, with depth, contact events and frame-accurate action segments aligned on one hardware clock where the spec includes them.

The broad egocentric data center page covers all twelve rows in the environment in a paragraph each. This page narrows to the two rows that mount hardware to the rack itself, so a spec for rail and panel work can be read in one place. The data is human capture by paid, consented contributors, not robot or teleoperation data, so use for a specific robot still needs your own retargeting step.

Catalog rows

Which catalog rows cover rack rail and cage nut mounting, and what is their status?

Two, both listed under the Data Center & IT Hardware Bench environment in the Office & IT sector, whose examples read "Server rack plus hardware bench: modules, cabling, drivers". "Screw-mount rack rails" is described as "Fasten rail brackets with a driver." and is rated Easy. "Clip cage nuts + panel" is described as "Clip cage nuts and mount a blanking panel." and is rated Medium.

Both are listed as Available on request, with the capture setup "Multiview rig + handheld grippers". Available on request means the work is scoped and recruited to spec when a buyer asks for it, rather than running as an open capture stream today. The site states no delivered hours for either row, so this page claims none.

Difference

How is this different from the other data center pages?

The egocentric data center and IT hardware page is the broad page: all twelve rows, from cable dressing to asset tags, with these two rows listed among them. The egocentric server hot-swap and component seating page covers five other rows, and it lists "Clip cage nuts + panel" and "Screw-mount rack rails" as rows it does not cover. The egocentric network patching and fiber transceiver page also lists them as rows outside its scope.

The stereo data center rack and server hardware page treats all twelve rows from the stereo and container side. None of them has rail and panel mounting as its main topic, which is why this page exists. The comparison table below sets the pages side by side.

Rail brackets

How is the rail row described?

One row, "Screw-mount rack rails", and one sentence: "Fasten rail brackets with a driver." It is the only mounting row that names a tool, a driver, and the site states no driver model, bit, torque or fastener type. The row is rated Easy, which is all the site says about its difficulty.

Rail length, bracket type, rack standard, number of brackets per episode and what a correctly fastened bracket looks like are Not stated. A spec that depends on any of them has to name them, and the accept criteria say what counts as a finished step.

Cage nuts and panel

How is the cage nut and blanking panel row described?

One row, "Clip cage nuts + panel", described as "Clip cage nuts and mount a blanking panel." It is the one Medium row of the two. The description holds two parts in one row, the cage nuts and the panel, so a spec can decide whether an episode covers both parts or one.

Cage nut counts, panel size, the order of the steps, hole patterns and any hand tool used for clipping are Not stated. The catalog does not say whether the two parts are separate steps or one, and it gives no way of labelling or measuring a mounted result.

Why demanding

Why is rack mounting work worth capturing for robot learning?

The catalog describes steps that end in a physical fit to the rack: brackets fastened with a driver, cage nuts clipped and a panel mounted. Lighting, rack density, reach and the height of the work are Not stated. The two rows are rated Easy and Medium, and this page says no more about difficulty than that.

What the site does state is what egocentric video puts into the training distribution: hands entering the frame from below, occlusion during grasp and motion blur during reach. RGB alone leaves a model to infer 3D hand structure it cannot see reliably through occlusion and motion blur, which is why the site delivers hand pose with a visibility flag per joint where a spec includes it. Whether that stream is populated for the staged gripper capture mode is Not stated.

Capture setup

How is rack mounting work captured, and is it first-person or stereo?

The catalog describes the mode 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 does not list a head-mounted phone or a stereo pair for either row, and this page does not claim either is running on them. The site states no camera count, resolution or frame rate for these rows, so the capture tier is agreed in the spec.

Stereo applies only through the site’s general stereo offer: a stereo pair is a capture choice agreed in the skill spec and confirmed on a Spec Pilot, and the stereo data center page covers that side. Whether the first-person head-mounted mode listed for other rows is offered for these two is Not stated. The site states no gripper model or jaw type, and the setting is agreed in the spec, since the site lists capture at a customer’s own facility as not offered.

Streams and labels

What does each rack mounting episode contain?

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. For mounting work, contact events mark when a bracket, a cage nut or a panel meets the rack, and action segments mark each step as the skill spec defines it.

The site does not state which of those streams the staged gripper mode populates for each row. Object labels use the terms your spec sets. Force or tactile data, and success or quality labels for a fastened bracket, a clipped nut or a seated panel, are Not stated. The streams table below sets each against what the site states.

Gloves

Can tactile gloves be used for rack mounting work?

The site offers a tactile glove tier as a pilot-ready option. The glove signal is time-synced to a head-mounted stereo camera by hardware timestamps and an IMU, and each glove is calibrated every session.

The site does not state that the glove tier has been used on either mounting row, and both rows list handheld grippers as their capture mode. Treat force and tactile data as Not stated for these rows unless a spec adds the glove tier and agrees the fit with you.

Setting and privacy

Where is it captured, and how are faces, screens and labels handled?

The catalog places the rows in a server rack plus hardware bench environment. The site names no data center, colocation facility, customer or operator, and it does not state capture at a customer’s own site. Environment, access and schedule are agreed in the spec, and the number of sites is Not stated.

Every episode links to a signed contributor release through a documented consent chain. Faces, plates and screens are detected and irreversibly blurred at ingest, a reviewer confirms the result before anything ships, and the anonymization record is delivered with the data. An undocumented bystander is a stated reject criterion. How asset tags, serial numbers, barcodes and cable labels are handled is Not stated, so name them in the spec.

Quality gates

How are rack mounting episodes checked for lost views?

Episodes are machine-checked at ingest against the stated reject criteria, listed in the failure-mode table below, and a human reviewer then checks the result. Two stated rules bite on this work: an episode is rejected when the manipulated object is out of frame for more than 15 percent of it, and when more than 35 percent of depth pixels in the manipulation region are invalid.

What counts as the object, a bracket, a cage nut or a panel, is set in the spec. How often a part leaves the view in a rack, how depth behaves on rack metalwork, and any blur or sharpness rule are Not stated. A rejected item goes to the reject log and is recaptured at our cost.

Delivery

How is rack mounting data labeled and delivered?

Delivery formats are RLDS, LeRobot and WebDataset, plus HDF5, zarr and Rerun .rrd. Firsthand also offers MCAP delivery; the MCAP hub page covers the container as its own documentation describes it, and which streams would be channels in a file, and what they are named, is Not stated and is agreed in the skill spec.

Every batch is validated against signed accept criteria before it ships, with a reject log, and delivery is rolling, per validated batch. The formats table below sets each format against what its official documentation states.

Scale and speed

How do custom, fast and on-demand collection apply to rack rail and cage nut mounting?

None of the site’s off-the-shelf dataset descriptions names rack rail or cage nut mounting, so for this task it 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 operators need to be recruited and vetted first. Both rows are Available on request, so which range applies is confirmed when the spec is quoted rather than assumed. On-demand top-ups run against the same spec and accept criteria after each training run. Collections are billed per validated hour, and the site publishes no price list, minimum order or batch size for these rows.

Ordering

How do you order a rack rail and cage nut mounting dataset?

Describe which of the two rows are in scope, where each task starts and ends, whether an episode covers one part or the whole step, and what counts as finished. Add the setting, the privacy rule for screens and labels, and the variation your pipeline is missing. Then name the streams you want, the capture mode, 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.

At a glance

What you get.

Catalog row 1
"Screw-mount rack rails": "Fasten rail brackets with a driver." Data Center & IT Hardware Bench, Office & IT; Easy; Available on request
Catalog row 2
"Clip cage nuts + panel": "Clip cage nuts and mount a blanking panel." Data Center & IT Hardware Bench, Office & IT; Medium; Available on request
Catalog capture setup
Multiview rig + handheld grippers, a staged working set where the operator performs the task with handheld grippers
Episode schema
Ten streams synced under 2 ms; which streams the gripper mode populates is Not stated
Delivery
RLDS, LeRobot, WebDataset, HDF5, zarr, Rerun .rrd, MCAP or raw streams with metadata
Tool models, fastener types, torque, rack dimensions, load ratings
Not stated; agreed in the skill spec
License
Perpetual, irrevocable, buyer-owned, non-exclusive by default

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 rows

Which catalog rows cover rack rail and cage nut mounting?

The two rows relied on by this page, reproduced as the catalog states them, plus items the catalog does not list.
Task (as listed)Description (as listed)EnvironmentSkillCapture setupStatus
Screw-mount rack railsFasten rail brackets with a driver.Data Center & IT Hardware BenchEasyMultiview rig + handheld grippersAvailable on request
Clip cage nuts + panelClip cage nuts and mount a blanking panel.Data Center & IT Hardware BenchMediumMultiview rig + handheld grippersAvailable on request
Racking a full server, configuring a switch, decommissioning hardwareNot stated as catalog rowsNot statedNot statedNot statedNot stated

Compared pages

How does this page differ from the other data center pages?

Each page, its main topic and how it treats the two mounting rows. “Not covered” means the page leaves them out of scope.
PageMain topicCapture framingTreatment of the two mounting rows
This page: egocentric rack rail and cage nut mounting dataThe two mounting rows onlyMultiview rig + handheld grippers, as the catalog lists itThe whole page: scope, limits, streams and ordering
Egocentric data center and IT hardware dataAll twelve rows in the environmentAs the catalog lists itTwo of twelve rows, described in short
Egocentric server hot-swap and component seating dataFive rows that swap or seat a partAs the catalog lists itNot covered; named as rows outside the page
Egocentric network patching and fiber transceiver dataPatching and transceiver rowsAs the catalog lists itNot covered; named as rows outside the page
Stereo data center rack and server hardware dataAll twelve rows, stereo and MCAP sideStereo pair agreed in the skill specTwo of twelve rows, with the others
Hard task data collectionEvery catalog row rated HardAs each row lists itNot covered; these rows are rated Easy and Medium

Stated vs not stated

What does the site state about capturing these two rows?

Left to right: the item, what the site states, and what that means for an order. “Not stated” means the site does not say.
ItemWhat the site statesFor a rack mounting order
Catalog capture setupA staged working set recorded by a multiview rig while the operator performs the task with handheld grippersStated for both rows; no head-mounted phone or stereo pair is listed for them
StereoA stereo pair is a capture choice agreed in the skill spec through the general stereo offerNot running on these rows as far as the site states; agreed in the spec
Camera count, resolution, frame rate for these rowsNot statedCapture tier agreed in the spec
Gripper model and jaw typeNot statedAgreed in the spec
Hand pose in gripper modeNot statedAgreed in the spec and confirmed on a Spec Pilot
Driver model, torque, fastener typeA driver is named in one row; the rest is Not statedNot claimed; set in the spec
Rack standard, rail and panel sizes, cage nut countsNot statedNot claimed; set in the spec
Faces and screensDetected and irreversibly blurred at ingest; reviewer confirms; footage that cannot be de-identified is discardedStated; record ships with the data
Asset tags, serial numbers, labelsNot statedName them in the spec
Row statusAvailable on requestNo volume, hours or delivered footage is claimed on this page

Streams and tasks

How do the stated streams map to mounting tasks?

Streams the site states for custom collections, set against what each carries in a mounting episode. Whether the gripper mode populates each is Not stated.
Stream or labelStatus on this siteWhat it carries in a mounting episode
Head and wrist RGB, depthStated; ten streams synced under 2 msKeeps the rack, the part and the hands on one timeline
3D hand pose, per-joint visibilityStated; two hands of 21 joints, with a visibility flag per jointCarries how the fingers hold a bracket, nut or panel and which joints were hidden
Contact and grasp eventsStatedMarks when a bracket, cage nut or panel meets the rack
Action segmentsStated; frame-accurateSeparates the fastening, clipping and mounting steps
Per-frame instance masksStated; stored at 848×480Follows the part set as the object in the spec
Object labelsSet by your specNames the parts and steps in the terms your policy uses
Force or tactile dataGlove tier stated; use on these rows Not statedAdded only if a spec includes the glove tier

Failure modes and QA

What goes wrong in rack mounting capture, and what does QA do about it?

Left to right: the failure mode, what the site states, and what a buyer sees. “Not stated” means the site does not say.
Failure modeWhat the site statesWhat a buyer sees
Fingers hidden behind a part or the driverPer-joint visibility flag on hand pose where a spec includes itA hidden joint flagged as hidden, not guessed
Part out of frameEpisode rejected when the manipulated object is out of frame for more than 15 percent of itReject log per batch; what the object is gets set in the spec
Invalid depth in the manipulation regionEpisode rejected when more than 35 percent of depth pixels are invalid; behaviour on rack hardware Not statedReject log per batch
Sync breachRejected beyond the ceiling for more than three consecutive framesReject log per batch
Calibration driftRejected beyond toleranceCalibration shipped with the episode
Bystander in frameUndocumented bystander rejected; faces and screens blurred and reviewer-confirmedReject log per batch; anonymization record ships with the data
Spec mismatchRejected for a mismatch against the named skill specReject log per batch
Blur or sharpnessNot statedSet in the spec if your policy needs a rule

Formats

Which formats carry rack mounting data?

Delivery formats stated on the site, set against what each format’s official documentation states where the documentation was read.
FormatWhat the official documentation statesUsed for here
RLDSA dataset is retrieved as episodes, and each episode holds a dataset of steps (RLDS repository README)Episode-based training loaders
LeRobotLeRobotDataset v3.0 is a standardized format for robot learning data, including multi-camera video, with many episodes stored per Parquet or MP4 file (LeRobot documentation)LeRobot dataset format and tooling
WebDatasetTar files in which files that belong together share a basename, split into numbered shards (WebDataset README)Sharded streaming for large training runs
MCAPAn open source container file format for multimodal log data, supporting multiple channels of timestamped pre-serialized data (MCAP site)Multi-stream files; topics and schemas Not stated, agreed in the spec
HDF5 and zarrStated by Firsthand onlyArray storage for custom pipelines
Rerun .rrdStated by Firsthand onlyMulti-stream synchronized playback for review

Delivery and licensing

What is delivered, and under what license?

Delivery and licensing terms as stated on this site. “Set per program” means the site states no fixed value.
ItemAs stated on this siteNote
FormatsRLDS, LeRobot, WebDataset, HDF5, zarr, Rerun .rrd, MCAP, or raw streams with metadataChosen in the skill spec; converter source ships with the data
LicensePerpetual, irrevocable, worldwide, buyer-owned, non-exclusive by defaultExclusivity is available as a separate term; the executed agreement controls
BillingPer validated hourRejected items are recaptured at our cost
PrivacyFaces, plates and screens blurred irreversibly; footage that cannot be de-identified is discardedAnonymization record delivered with the data
Timelinea median of 48 hours to first validated batch where coverage is deep; three to four weeks where thinWhich applies to these rows is Not stated; quoted against your spec
Price, minimum order, batch sizeNot published for these rowsSet in your quote

Skill spec

What does the skill spec fix for a rack mounting 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
Rows in scopeRail brackets, cage nuts and panel, or bothWhat the cameras see; set per program
Steps and orderWhat one step is and how steps are namedEpisode boundaries and action segments; set per program
Episode scopeOne part, one row or a longer runSet per program; the site does not state an episode length
Object definitionWhether a bracket, a cage nut, a panel or the driver counts as the objectContact events and the out-of-frame reject rule; set per program
Capture mode and streamsMultiview rig with handheld grippers, any stereo pair, wrist views and hand poseThe channels you can read; confirm on the Spec Pilot
People and privacyWhat may be in frame and how screens and labels are handledSettled before capture; faces and screens blurred at ingest
Accept criteriaWhat counts as a finished step and what passes QAReject log per batch
Delivery formatHow the data is packagedRLDS, LeRobot, WebDataset, HDF5, zarr or MCAP
LicenseWho owns the datasetPerpetual, irrevocable, buyer-owned, non-exclusive by default

Example footage

What does egocentric capture look like?

One real clip already published on this site, shown as a format and rig example: a head-mounted stereo capture with hands in frame. It does not show rack rail mounting, cage nuts or a blanking panel, and it is not a multiview-rig or handheld-gripper clip.

egocentric_stereo_sample.mp4 · head-mounted stereo · hands in frame · 1920 × 1080 web preview · format example, not a rack clip

No rack mounting clip is published on the site, so this is labelled as a format and rig example. Request a sample to see delivered files. See stereo sample page, request an egocentric data sample.

FAQ

Egocentric server rack installation data, answered.

What is egocentric rack rail and cage nut mounting data?
Real-world video of a paid, consented contributor fastening rail brackets with a driver, and clipping cage nuts and mounting a blanking panel, recorded against a written spec.
Which catalog rows cover it, and what is their status?
Two rows in the Data Center & IT Hardware Bench environment: "Screw-mount rack rails" ("Fasten rail brackets with a driver.", Easy) and "Clip cage nuts + panel" ("Clip cage nuts and mount a blanking panel.", Medium). Both are Available on request. This page claims no delivered hours.
Is this head-mounted or stereo capture?
The catalog lists a multiview rig with handheld grippers for both rows, a staged working set where the operator performs the task with grippers. It does not list a head-mounted phone or a stereo pair for them. A stereo pair is agreed in the skill spec through the general stereo offer.
How is this different from the egocentric data center page and the hot-swap page?
The egocentric data center page covers all twelve rows, and the hot-swap page covers five other rows and lists these two as outside its scope. This page covers rail and panel mounting only.
Does the site state tool models, torque, rack sizes or part counts?
No. A driver is named in one row, but tool models, torque, fastener types, rack standards, rail and panel sizes and cage nut counts are Not stated and are agreed in the spec.
How are faces, screens and labels handled?
Faces, plates and screens are detected and irreversibly blurred at ingest, a reviewer confirms the result, and the anonymization record ships with the data. How asset tags, serial numbers and labels are handled is Not stated, so name them in the spec.
How fast can a custom rack dataset be collected?
From signed spec to first validated batch, a median of 48 hours where coverage is deep and three to four weeks where it is thin. Both rows are Available on request, so ask which applies when you request the quote. Collection runs at scale through a network of thousands of capture providers, for a consistent, recurring data flow to labs.
Which formats can it be delivered in?
RLDS, LeRobot, WebDataset, HDF5, zarr, Rerun .rrd or raw streams with metadata, and MCAP on request. Topic names and schemas for MCAP are Not stated and are agreed in the skill spec.

Sources

Where are the formats documented?

Request a custom rack rail and cage nut mounting dataset.

Tell us which of rail bracket fastening, and cage nut and blanking panel mounting are in scope, where each task starts and ends, the capture mode, the privacy rule for screens and labels, and the streams and format you want. We will quote reach, timeline and price against a written spec.