Depth at the port and the transceiver cage
Depth is recoverable from every frame pair, so where a connector sits against a port, or a transceiver against its cage, is in the same capture as the hands. No depth accuracy figure is stated.
Now recruiting capture teams worldwide. Get paid to record first-person video of real work.
Apply to captureTask family
Why stereo and MCAP here
Depth is recoverable from every frame pair, so where a connector sits against a port, or a transceiver against its cage, is in the same capture as the hands. No depth accuracy figure is stated.
A hand covers the port it is working at. Hand pose carries a per-joint visibility flag, so a hidden joint is flagged, not guessed.
Streams are hardware-triggered onto one clock with a 2.0 ms cross-stream ceiling, and action segments can mark each step as the skill spec defines them.
Definition
It is first-person video of a person doing the two network tasks the catalog names, recorded from their own point of view with two calibrated cameras and with hand pose and sensor streams alongside. Stereo describes the capture, and MCAP describes the container. They are separate choices.
The site does not state that stereo is already running on these rows. The catalog lists a multiview rig with handheld grippers for both, and a stereo pair is a capture choice agreed in the skill spec and confirmed on a Spec Pilot. The data is human capture by paid, consented contributors, not robot or teleoperation data. Firsthand is not affiliated with Foxglove or the MCAP maintainers.
Difference
The egocentric network patching page covers the same two rows as first-person task data, with a stream map, reject criteria and delivery formats, and it states that it is not a stereo page. The stereo data center page covers the stereo and container side of all twelve rows in the environment, so patching and transceivers are two of twelve there.
This page is the stereo and MCAP treatment of just the two network rows and the failure modes they share. It states no switch, transceiver or cable model, no port count, no link speed and no standard beyond the catalog wording, because the site states none.
Catalog rows
Two core rows, both in the Data Center & IT Hardware Bench environment under the Office & IT sector, which the catalog describes as a server rack plus a hardware bench with modules, cabling and drivers. "Patch RJ45 cabling" is rated Easy and "Insert SFP+ + fiber" is rated Medium. Both list a multiview rig plus handheld grippers, and both are Available on request.
Available on request means the site does not state that either row has been collected and delivered. This page claims no delivered hours and no active collection. Two nearby rows, "Cable-comb dressing" and "Connect C13/C14 power", also name a cable, but they are not network rows and are context only.
Ethernet
The catalog description of "Patch RJ45 cabling" reads "Insert and route Ethernet patch cables cleanly." Two verbs carry the row: insert, which ends at a port, and route, which follows the cable along its path. The word "cleanly" and the connector name RJ45 are the catalog wording.
The site states no port count, cable length, cable type, patch panel layout or link speed for this row, and no definition of a clean route. A spec that depends on any of them has to name them.
Fiber
The catalog description of "Insert SFP+ + fiber" reads "Seat a transceiver and connect an LC fiber." It is rated Medium, one of three Medium rows among the twelve in the environment. The row has two steps in its name: seat the transceiver, then connect the fiber.
The catalog gives no way of labelling or measuring a seated transceiver or a connected fiber. Module counts, port counts, fiber types and the order of the two steps are Not stated, and are settled in the written spec.
Not stated
The site says tasks people often expect in this topic, such as configuring a switch, testing a link or decommissioning hardware, are not catalog rows. Tasks outside the list can be added to a spec, but this page claims no coverage of them.
Eight other rows in the same environment are not network rows and are not covered here: asset tags, cage nuts and panels, rack rails, drive caddies, PSU and fan modules, RAM DIMMs and PCIe cards. They are on the stereo data center page.
Why stereo
Network work is about where a small connector meets a port: the line of a plug against a socket, the position of a transceiver against its cage, the path of a cable after it is inserted. A stereo pair recovers depth from the calibrated frame pair rather than from a standalone depth sensor, so the form of the port and the connector is in the same frame as the hands.
No depth accuracy figure, connection label or pass or fail field is stated for these rows.
Capture setup
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." Whether hand pose is recorded while handheld grippers are used is Not stated and is agreed in the spec, as is the gripper model and jaw type.
The site describes two stereo setups and does not say which one these rows use. The hard-environments stack is a global-shutter 4K/60 head camera with a 120° lens, two 1080p/60 wrist cameras, a depth sensor at 848×480 and 30 fps, and 200 Hz 6-DoF IMUs. The off-the-shelf stereo head rig, dataset 06, is dual fisheye cameras at 1080p per eye, a 62 mm baseline, per-unit calibration, a 9-axis IMU and hand-tracking output. How they combine with the multiview rig is Not stated.
Failure modes
The first risk is occlusion by the hand itself: a hand covers the port while the connector is inserted. The site answers with a per-joint visibility flag, contact events, and wrist views where the spec asks for them.
Depth is the second. The site says depth degrades on wet and reflective surfaces, masks low-confidence regions with per-frame confidence maps, and rejects an episode above 35 percent invalid depth pixels in the manipulation region. How depth behaves on rack metalwork, transceivers or fiber connectors is Not stated. Calibration is the third, measured at the start and end of every session under a 0.28 px RMS gate. An episode is also rejected when the manipulated object is out of frame for more than 15 percent of it or an undocumented bystander appears, and a reject log ships with every batch. The site states no handling for glare, low light or motion blur on these rows, so this page promises none.
MCAP
MCAP is an open source container file format for multimodal log data, supporting multiple channels of timestamped pre-serialized data. It is serialization-agnostic, so messages can be Protobuf, DDS (CDR), ROS, JSON and more. Its design is row-oriented and append-only, it stores schemas alongside the data, and it offers an optional index plus optional LZ4 or Zstandard chunk compression.
Reader and writer libraries exist for C++, Go, Python, Rust, Swift and TypeScript. The foxglove/mcap repository is MIT-licensed; that is the MCAP project’s licence, not the licence of a Firsthand delivery.
Streams
Firsthand offers MCAP delivery, and dataset 06 is stated as shipping as MCAP. For patching work the site states hand pose, a stereo-derived depth stream, IMU, contact events, action segments, per-frame instance masks and one task instruction per episode elsewhere. What it does not state is which of those are channels in an MCAP file, how they are named, what encoding they use or at what rate.
No topic name, message schema or rate is stated here; confirm them in the skill spec on a Spec Pilot batch.
Skill rating
The Skill column lists "Patch RJ45 cabling" as Easy and "Insert SFP+ + fiber" as Medium, and that is all the site says about difficulty. Medium is shared by three of the twelve rows in the environment. This page adds no rubric, failure rate, tolerance or claim about worker skill, because the site states none.
What the row text does say is what each step involves: insert and route for the Ethernet row, and seat then connect for the transceiver row. Both end with a small connector meeting a port, and the cable routing page states the general problem plainly: a connector has a small target, and the hands do the work that decides whether it ends up where it belongs.
First-person view
The site states that egocentric video puts hands entering the frame from below, occlusion during grasp and motion blur during reach into the training distribution. 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.
A stereo pair adds recoverable depth to the same frames, so the connector, the port and the cage sit in three dimensions beside the hands. What it does not add is a label: the site states no connection label and no seating label for either row.
Streams and labels
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. Hand pose is delivered as handJoints[2][21][3] in metric units in the head-camera frame, aligned to depth and RGB.
For patching work, contact events can mark when a connector meets a port and action segments can mark each step. 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, and force or tactile data is Not stated.
Gloves
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 network row, and both list handheld grippers as their capture mode. Treat force and tactile data as Not stated unless a spec adds the glove tier and agrees the fit with you.
Quality gates
Episodes are machine-checked at ingest against the stated reject criteria, and a human reviewer then checks the result. The criteria are an object out of frame for more than 15 percent of the episode, more than 35 percent invalid depth pixels in the manipulation region, a sync breach beyond the ceiling for more than three consecutive frames, calibration drift beyond tolerance, an undocumented bystander, and a mismatch against the named skill spec.
The site states no rule for how long a port may stay hidden behind a hand, and no blur or sharpness rule. A rejected item goes to the reject log and is recaptured at our cost.
Formats
Delivery formats are RLDS, LeRobot and WebDataset, plus HDF5, zarr and Rerun .rrd. By their official documentation, RLDS retrieves a dataset as episodes with each episode holding a dataset of steps, LeRobotDataset v3.0 is a standardized format for robot learning data including multi-camera video with many episodes per Parquet or MP4 file, and WebDataset uses tar files in which files that belong together share a basename, split into numbered shards.
Delivery is rolling, per validated batch, with a reject log. Whether MCAP is delivered alone or beside one of these is set in the spec.
Scale
Off-the-shelf covers what is already in stock. The industrial and skilled-trade category lists datasets 02, 03 and 04, and the stereo category lists the stereo head rig. Neither listing names data center, network patching or fiber work, so this page claims no off-the-shelf coverage of these rows.
Custom collection covers the rest: custom collection at scale through a network of thousands of capture providers, for a consistent, recurring data flow to labs. The site states no provider count, cadence, hours or volume beyond that wording, so they are quoted against your spec.
Fast · on demand
Where coverage is deep, the median time from signed spec to first validated batch is a median of 48 hours. Where coverage is thin and contributors need to be recruited first, a first batch typically takes three to four weeks. Both rows are Available on request, and the site does not say which timing applies to them, so ask when you request the quote.
On-demand top-ups run against the same spec and accept criteria after each training run.
Privacy and rights
For every delivered episode the site states that faces, heads, licence plates and legible screens are detected and irreversibly blurred at ingest, a human reviewer confirms the result, and footage that cannot be de-identified is held or discarded. Every episode links to a signed contributor release through a documented consent chain, and the release covers bystanders. How cable labels, port labels and asset tags are handled is Not stated, so name them in the spec.
The buyer owns the dataset under a perpetual, irrevocable, non-exclusive-by-default licence, with exclusivity available. The site names no data center, colocation facility, customer or operator, and this page makes no data center access, safety or certification claim for any row.
Ordering
Name the patching and transceiver tasks, the steps in order, what the camera must see at the port, and the streams, topic naming and encoding your reader expects. Say whether MCAP is the only delivery or sits beside RLDS, LeRobot or WebDataset. Most programs begin with a Spec Pilot, which confirms the spec on real footage, and every batch is validated against signed accept criteria before it ships.
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 rows
| Task (as listed) | Description (as listed) | Skill | Capture setup | Status |
|---|---|---|---|---|
| Patch RJ45 cabling | Insert and route Ethernet patch cables cleanly. | Easy | Multiview rig + handheld grippers | Available on request |
| Insert SFP+ + fiber | Seat a transceiver and connect an LC fiber. | Medium | Multiview rig + handheld grippers | Available on request |
| Cable-comb dressing (nearby, context only) | Dress a cable bundle to uniform pitch with ties. | Easy | Multiview rig + handheld grippers | Available on request |
| Connect C13/C14 power (nearby, context only) | Seat power cables and locking clips. | Easy | Multiview rig + handheld grippers | Available on request |
Neighbouring pages
| Page | Rows it covers | Stereo and MCAP content |
|---|---|---|
| Egocentric network patching and fiber transceiver data | The same two rows: "Patch RJ45 cabling" and "Insert SFP+ + fiber" | Egocentric treatment with stream map and reject criteria; states it is not a stereo page; MCAP is described on the stereo data center page |
| Stereo data center rack and server hardware data | All twelve Data Center & IT Hardware Bench rows, patching and transceivers among them | Stereo and MCAP treatment across all twelve rows |
| Egocentric data center and IT hardware data | All twelve rows as task video data | Stereo named only for the tactile glove tier; no MCAP content |
| Stereo thermal paste and heatsink application data | One row in Precision Assembly Bench (electronics): "Apply paste + heatsink" (Medium) | Stereo and MCAP treatment for a different environment and task; not a network page |
| This page | The two network rows only | Yes: stereo and MCAP treatment of patching and transceiver work |
Capture setup
| Item | What the site states | For patching and transceiver rows |
|---|---|---|
| Catalog capture setup | Multiview rig + handheld grippers | Listed on both rows |
| Stereo head rig (06) | Dual fisheye, 1080p per eye, 62 mm baseline, per-unit calibration, 9-axis IMU, hand-tracking output; shipped as MCAP | Agreed in the skill spec through the general stereo offer; not listed on the rows |
| Hard-environments stack | Global-shutter 4K/60 head camera, 120° lens, two 1080p/60 wrist cameras, depth sensor 848×480 at 30 fps, 200 Hz 6-DoF IMUs | Which setup these rows use is Not stated |
| Hand pose with handheld grippers | Not stated | Agreed in the skill spec |
| Gripper model and jaw type | Not stated | Agreed in the skill spec |
Failure modes
| Failure mode | What the site says | Where it shows up in network work |
|---|---|---|
| Hand covers the port or cage | Hand pose carries a per-joint visibility flag; contact events mark when a hand meets an item; wrist views can be specified | Patch RJ45 cabling, Insert SFP+ + fiber |
| Wet or reflective surfaces degrade depth | Per-frame confidence maps mask low-confidence regions; an episode is rejected above 35 percent invalid depth pixels in the manipulation region | Depth behaviour on rack metalwork, transceivers and fiber connectors: Not stated |
| Calibration drifts over a session | Calibrated at the start and end of every session; 0.28 px RMS gate | Close work on a port or a cage |
| Clocks drift between streams | Hardware-triggered onto one clock; 2.0 ms ceiling | Pairing hand pose against the stereo pair across insert, route, seat and connect |
| Object out of frame, or an undocumented bystander | Episode rejected; reject log ships with every batch | A connector or fiber leaving view; someone entering frame |
| Glare, low light, motion blur | Not stated for these rows | Agreed in the skill spec |
| Connected or seated result | No connection label or measure of a clean route is stated | Patch RJ45 cabling, Insert SFP+ + fiber |
Stream map
| Stream or item | What the site states | In an MCAP delivery, as the site states it |
|---|---|---|
| Stereo pair | Dual fisheye, 1080p per eye, 62 mm baseline (dataset 06) | Dataset 06 is stated as shipped as MCAP; channel layout not stated |
| Depth with confidence maps | Recovered from the calibrated stereo pair; confidence map masks low-confidence regions | Not stated |
| Hand pose | 21 joints per hand with a per-joint visibility flag | Channel layout and joint order not stated |
| IMU | 9-axis (dataset 06); 6-DoF at 200 Hz (hard-environments stack) | Channel layout and rate not stated |
| Contact events, action segments, instance masks | Listed in the site’s episode schema; masks stored at 848×480 | Not stated |
| Task instruction | One per episode | Placement in the file not stated |
| Calibration | Measured every session and shipped with the episode | Placement in the file not stated |
| Topic names, schemas, encoding, rates | Not stated | Agreed in the skill spec |
Sources of data
| Question | Off-the-shelf catalog | Custom collection at scale |
|---|---|---|
| What it covers | Datasets already captured; the industrial category names machinery, auto repair, carpentry, construction, factory floors and inspection | Tasks, setting and streams you write into a skill spec |
| Network patching and fiber tasks | Not named in the listings | Two catalog rows, both Available on request |
| Time to first data | In stock to license | a median of 48 hours where coverage is deep; three to four weeks where it is thin |
| Supply model | Not applicable | A network of thousands of capture providers, for a consistent, recurring data flow to labs |
| Provider count, cadence, hours, volume | Not stated | Not stated; quoted against your spec |
| License | Non-exclusive, with resale rights | Perpetual, irrevocable, buyer-owned, non-exclusive by default |
Example footage
One real clip already published on this site: a head-mounted stereo capture with hands in frame, shown as a rig and format example. It does not show patching a cable or seating a transceiver.
No network patching or fiber transceiver clip is published on the site, so this is the nearest clip. Request a sample to see delivered files. See stereo sample page, stereo + wrist-cam example videos, request an egocentric data sample.
FAQ
Sources
Tell us the patching and transceiver tasks, the steps in order, what the camera must see at the port, the streams you want in the file and whether MCAP sits alone or beside a training format. We will quote reach, timeline and price against a written spec, with a first validated batch in a median of 48 hours where coverage is deep.