Stereo Bicycle and Appliance Repair Data

It is first-person video of a person doing hands-on repair work, recorded with two calibrated cameras, with hand pose and sensor streams alongside. People reach it through several phrasings: bicycle repair robot training data, stereo appliance repair video, an egocentric repair shop dataset, or custom repair data collection at scale. Stereo describes the capture, and MCAP describes the container. They are separate choices.

It is task video, not a repair manual. The data is human capture by paid, consented contributors, not robot or teleoperation data, so using it for a specific robot still needs your own retargeting step. The site states one task instruction per episode as its language data. This page gives no repair, electrical or appliance-safety instruction. Firsthand is not affiliated with Foxglove or the MCAP maintainers.

Two, listed in lib/commercial-catalog-data.ts under the Repair Shop environment, in the Repair Services sector. "Bicycle repair" reads "Adjust brakes and gears, replace tubes, chains and cables and true wheels" and is rated Medium. "Appliance repair" reads "Open the appliance, test components, replace the faulty part and reassemble" and is rated Hard. Both list "Phone (head-mounted) / multiview rig" as the capture setup. The first table below reproduces them with the catalog wording.

Both are Available on request, which means the work is scoped and recruited to spec when a buyer asks for it. Neither is listed as Active collection or Ready to deploy, so the site does not say a bicycle or appliance capture stream is running today. The site states no delivered hours and no delivered stereo footage for these rows, so this page claims none.

This page is the stereo and MCAP treatment of both rows together. The egocentric bicycle repair page covers the Bicycle repair row, and the egocentric appliance repair page covers the Appliance repair row, each as first-person video that is not stereo as its subject. Their capture setup is the same head-mounted phone or multiview rig.

The stereo electronics and device repair page covers a different row, Device Repair, in a different environment, Electronics & Device Repair Shop, and that row is Active collection. It concerns phones, laptops, tablets and consumer electronics. The stereo handyman and home repair page covers four On-Site Home Repair rows, all Available on request, where the work is done in a client home. This page is stereo capture of bicycle repair (brakes, gears, tubes, chains, cables, wheel truing) and appliance repair (opening, component testing, part replacement, reassembly). The second table below sets the pages side by side.

The catalog lists "Phone (head-mounted) / multiview rig" for both rows. The site does not state a stereo pair for either of them. 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. Camera count, resolution and frame rate for repair work are Not stated.

The site describes two stereo setups and does not say which one an order for these rows uses. 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. On the rig tier every stream is hardware-triggered onto one PTP clock domain, and calibration is solved before capture starts and after it ends.

The Bicycle repair row names six jobs: brakes, gears, tubes, chains, cables and wheels, written as adjusting brakes and gears, replacing tubes, chains and cables, and truing wheels. Those are the whole of what the site says a bicycle repair dataset covers. Other bicycle jobs, such as cleaning a chain or fitting accessories, are Not stated, so none is claimed. The row is rated Medium, the only row in its environment not rated Hard.

In stereo, a cable run, a wheel and the hands working them are recorded in the same frame pair. The egocentric bicycle page states that the hands often sit close to a small part that the head view only partly sees, and that a wrist-mounted camera is available alongside the head camera for close-range detail. Whether a bicycle is held in a stand or placed on the floor, and which bicycle types and stands are used, are Not stated. A spec that treats each job as its own task, with a start state, an end state and accept criteria, gives action segments that stay consistent across contributors.

The Appliance repair row names four steps: open the appliance, test components, replace the faulty part and reassemble. It is rated Hard. It does not name a diagnosis step, and it does not name a screen, battery or board, which belong to the separate Electronics repair row. Appliance types, faults, tools and test equipment are Not stated.

An opened appliance is an enclosed, cluttered volume where the hands, a tool and a small part share a small space. A calibrated pair records how those sit relative to each other, and action segments can separate opening, testing components, replacing the part and reassembling when the spec names them. The site gives no electrical or appliance-safety standard, no certification claim and no fault-diagnosis accuracy, and this page makes none. Whether a part can be seen while it is replaced is Not stated.

The Repair Shop environment lists examples that include an appliance workshop and a bicycle shop. Both rows use the capture mode in which a head-mounted phone records the first-person view of real work on site, and a multiview rig adds synchronized third-person angles when the spec asks for them. The site names no shop, brand or mechanic, and does not state the number of sites, the regions or the venue type. Environment, access and schedule are agreed in the written spec.

The site does not describe tool changes as a separate event. What it states is that contact events mark when the hands or a tool meet a surface or part, and that action segments mark each step. Bench layout, lighting and inspection routines for these rows are Not stated, so a spec that depends on any of them should say so and confirm it on a Spec Pilot.

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 or a tool meets a part, and why instance masks are recorded per frame. In repair, contact events mark touches on a brake, a cable, a wheel, a cover or a replaced part, and action segments mark each step.

The site says two-handed fine motion and small objects are where occlusion bites, and it states that depth degrades on wet and reflective surfaces. Whether bare metal, a wheel rim, a chain or an appliance panel counts as reflective is Not stated, and so is whether a small fastener leaves the view for long. A spec that needs a clear view of a step should name that step, ask for wrist views or a multiview angle, and confirm it on a Spec Pilot.

The site states a consent chain for contributors. Every contributor signs a written release, each delivered episode links to it through consent artefact IDs, and a break at any of the six checkpoints discards the footage. An undocumented bystander is a stated reject criterion. For every delivered episode the site states that faces, plates and screens are detected and irreversibly blurred at ingest, that a reviewer confirms the result before anything ships, and that footage that cannot be de-identified is discarded. The anonymization record ships with the data.

The site states no rule for a shop customer who is nearby, for an appliance owner, for a customer home or for data stored on a smart appliance, so how people who are present are handled is settled in the spec. This page describes capturing repair work, not customers. This is not legal advice, and the executed agreement controls.

Shop, client, bicycle and appliance brand and model names, electrical or appliance safety standards, torque specifications, repair times, error rates, bicycle and appliance types, camera count, resolution and frame rate for repair work, and what a correct or finished repair means are Not stated. They are set in the written spec, and this page claims none of them.

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, and the mcap CLI can inspect, merge and split files. The foxglove/mcap repository is MIT-licensed; that is the MCAP project licence, not the licence of a Firsthand delivery. The MCAP hub page and the glossary entry cover the format in full.

Firsthand offers MCAP delivery, and dataset 06 is stated as shipping as MCAP. For these rows the site states a stereo pair, hand pose, a stereo-derived depth stream with confidence maps, 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 on this page, and none should be assumed. Name the streams that matter in the skill spec and confirm them on a Spec Pilot batch. The fourth table below sets each stream beside what the site states about it.

Hand occlusion by the part, the tool or the frame of the bicycle or appliance is the first risk, under the visibility flag above. Small parts that leave the view are the second, and how often that happens is Not stated. Calibration is measured at the start and end of every session under a 0.28 px RMS gate, and an episode is rejected for a sync breach beyond the ceiling for more than three consecutive frames, for the manipulated object out of frame for more than 15 percent of the episode, for more than 35 percent of depth pixels in the manipulation region being invalid, for an undocumented bystander or for a mismatch against the named skill spec.

A reject log ships with every batch, and rejected items are recaptured at the provider’s cost. The fifth table below lists the failure modes beside what the site states.

The October 2026 off-the-shelf catalog lists ~324,800 hours across 18 datasets, of which stereo is ~64,800 hours across 9 datasets. None of them names bicycle repair or appliance repair, so for stereo capture of these tasks this 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. Where coverage is deep, the median time from signed spec to first validated batch is a median of 48 hours; where it is thin, turnaround runs three to four weeks. The site does not say which applies to these rows, and both are Available on request, so the timeline is quoted against your spec. Collections are billed per validated hour, and rejected episodes and recaptures are never billed.

Describe the jobs and where each starts and ends: a brake adjusted, a tube replaced, a wheel trued, an appliance opened and its components tested, a part replaced and the appliance reassembled. Say whether an episode covers one step, one job or a whole repair, what counts as the object and as finished, the setting and lighting, the privacy rule for the shop and for anyone in frame, and name the variation your pipeline is missing. Then name the streams you want in the file, the topic naming and message encoding your reader expects, whether calibration ships inside the file or beside it, and whether MCAP is the only delivery or sits beside RLDS, LeRobot or WebDataset.

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 sixth table below shows what each spec element decides.