Egocentric Bicycle Repair Video Data

Egocentric bicycle repair data is real-world video recorded from the mechanic point of view while a person works on a bicycle. The camera sits on the head, so the footage shows what the person sees: the hands, the part, the tool and the frame around them. Alongside the video, each episode carries depth, 3D hand pose, contact events and frame-accurate action segments, all aligned on one hardware clock.

Teams search for it under several names, including bicycle repair robot training data, a bike maintenance video dataset and egocentric bike mechanic data. They all describe first-person demonstrations of repair work, collected to a written spec. This page covers only what the Firsthand site states about it, and marks everything else as Not stated.

The site states bicycle work in one place, its commercial task catalog. A Bicycle repair row reads "Adjust brakes and gears, replace tubes, chains and cables and true wheels." That sentence names six jobs: brakes, gears, tubes, chains, cables and wheels. Those six are the whole of what the site says a bicycle repair dataset covers.

Other bicycle jobs are Not stated. That includes cleaning or lubricating a chain, fitting accessories and any full overhaul, 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 row sits in the Repair Services sector, in the environment called Repair Shop, covering appliance, bicycle, shoe, watch and furniture work. It is rated Medium, which makes it the only row in that environment that is not rated Hard. It is recorded with a head-mounted phone or a multiview rig, and its status is Available on request.

Available on request means the work is scoped and recruited to spec when a buyer asks for it. It is not listed as Active collection, so the site does not say that a bicycle capture stream is running today. A first batch is therefore quoted against your spec rather than drawn from a stream that already exists.

The Repair Shop environment lists a bicycle shop among its examples, next to electronics repair, appliance workshop, shoe repair, watch repair and furniture restoration. The bicycle row uses the capture mode in which a head-mounted phone records the first-person view of real work on site. That is the real-work mode. The site describes a different mode, a multiview rig with handheld grippers, as a staged working set, and the bicycle row does not use it.

The site names no shop, brand or mechanic, and does not state the number of sites, the regions or the type of venue. It does not state capture at a customer own facility either. Environment, access and schedule are agreed in the written spec.

A head-mounted phone records the first-person view of the job, and a multiview rig adds synchronized third-person angles when the spec asks for them. A wrist-mounted camera is available alongside the head camera for close-range detail. Both views are useful in bicycle work, where the hands often sit close to a small part that the head view only partly sees.

The site does not state a camera count, resolution or frame rate for repair. It also does not state whether a bicycle is held in a stand or placed on the floor, since bicycle types, stands and bench set-ups are Not stated. Raise any of them in the spec before capture starts.

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 bicycle repair, hand pose tracks the fingers on a brake or gear adjustment, contact events mark when the hands meet a part or tool, and action segments mark each step of the job.

Object labels use the terms your spec sets. Component types, tool types, torque values, repair durations, cycle times and number of episodes are Not stated. Success or quality labels, certifications and robot results are Not stated too. The site makes no safety-compliance claim for this work.

The catalog rates the row Medium, so the site does not call it Hard, and this page makes no claim about robot results. What the row does show is breadth. One line bundles six jobs, and each is its own sequence of hand steps. A brake adjustment, a cable replacement and a wheel truing job do not share a step order, so a policy trained on one tells you little about another.

The viewpoint matters too. The site states that a robot or humanoid perceives from a head- or wrist-mounted camera, not a static overhead shot, and that egocentric capture matches that geometry. Repair work is also close-range hand work, where the hands and the part block much of the view from a fixed camera. A first-person view keeps them in frame, which is why a spec for this kind of work asks for it.

A good spec treats each named job as its own task. Brake adjustment, gear adjustment, tube replacement, chain replacement, cable replacement and wheel truing each get a start state, an end state and a set of accept criteria. That keeps action segments consistent across contributors and makes it possible to ask for more of one job without re-collecting the others.

The spec also fixes what the site leaves open: bicycle types, component types, tool types and the order of steps inside a job. The task breakdown table below lists each job with the detail the site does not state.

The same Repair Shop environment holds five other rows: Appliance repair, Electronics repair, Furniture restoration, Shoe repair and Watch repair. All five are rated Hard and Available on request. The electronics and device repair page covers the electronics and appliance rows in depth, and this page does not restate them.

The vehicle maintenance and tire and wheel service pages cover other work, and neither states a bicycle task. The tool use page covers general tool handling. They are listed below as neighbours, not as sources for anything stated about bicycles.

Delivery formats are RLDS, LeRobot and WebDataset, plus HDF5, zarr and Rerun .rrd. Bicycle 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.

Not in standard capture. Force or tactile readings are Not stated for the bicycle row. The site separately offers a tactile glove tier as a pilot-ready option, in which glove signals are time-synced to a head-mounted stereo camera by hardware timestamps and an IMU, each glove is calibrated every session, and the pipeline checks signal dropout and sync drift.

The site does not state that the glove tier has been used on bicycle repair, or whether gloves suit fine adjustment work. If your policy needs a contact signal, ask for the glove tier in the spec and agree the fit with us first. The tactile glove page covers the tier in full.

Every episode links to a signed contributor release through a documented consent chain, and the consent record is part of the delivery. Faces, plates and screens are detected and irreversibly blurred at ingest, and a reviewer confirms the result before anything ships. The anonymization record is delivered with the data.

That is the whole of the privacy wording the site states. It does not state a rule for a shop customer who is nearby, and it does not state in-home capture, so this page makes no such claim. If either matters to your program, raise it in the spec and it is agreed before capture begins.

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 the bicycle row, and the row is 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 bicycle 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.