Stereo Weld Inspection and Cleanup Data

It is first-person video of post-weld work in a fabrication shop: inspecting welds for defects, grinding and cleaning seams, and marking parts for rework. It is recorded from the operator’s own point of view with two calibrated cameras, 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 this row. The catalog lists a head-mounted phone or a multiview rig as its capture setup, so 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, so using it for a specific robot still needs your own retargeting step. Firsthand is not affiliated with Foxglove or the MCAP maintainers.

One, in the Metal Fabrication & Welding Workshop environment of the Manufacturing & Workshops sector, listed in lib/commercial-catalog-data.ts. "Weld Inspection & Cleanup" is described as "Inspect welds for defects, grind and clean seams and mark parts for rework." It is rated Medium.

It is listed as Ready to deploy. The site states no delivered hours and no delivered stereo footage for it, so this page claims none. The environment examples read "Metal shop with welding bays, grinders, cutting and fitting benches"; these are kinds of workplace, not named shops, and this page names none. The first table below reproduces the row as the catalog states it.

This page is about post-weld inspection and cleanup in a fabrication shop. It is not welding itself, not general grinding and not generic surface QA. The stereo welding and metal fabrication page covers twelve rows across welding, layout, pipe, grinding and weld inspection, so the weld inspection row is one among many there. The stereo grinding and deburring page covers the two separate grinding rows of the same workshop. The stereo surface quality inspection page treats nine inspection rows across several environments, this row among them, and says no row is named surface quality inspection.

This page is the stereo and MCAP treatment of the one weld inspection row only, with its own stream map and failure modes. No weld type, defect type, defect rate, code, standard, NDT method or acceptance criterion appears here, because the site states none. The second table below sets the pages side by side.

The catalog lists "Phone (head-mounted) / multiview rig" for the row. The site does not state a head-mounted stereo setup for it. Stereo applies through the site general stereo offer: a stereo pair is a capture choice agreed in the skill spec and confirmed on a Spec Pilot.

A multiview rig adds synchronized third-person angles, and wrist cameras give a closer view of the hands, when the spec asks for them. Whether a stereo pair is worn, mounted or combined with a multiview rig on this row is Not stated, and so is where a rig sits relative to a weld bench or a grinder.

A seam is a feature of the shape of a part. A stereo pair recovers depth from the calibrated frame pair rather than from a standalone depth sensor, so the form of the seam and the part sits in the same capture that shows the hands working it. The site gives no depth accuracy figure for this row, so what a buyer does with that depth is their pipeline’s question.

A stereo pair adds depth, not a verdict. The site states 1080p per eye for dataset 06, and whether that resolves a defect in a seam at working distance is Not stated, so a spec should check it on a Spec Pilot.

The row begins with "Inspect welds for defects". It names the step and nothing about the method: the defect types, the inspection tools, any gauge or light, any non-destructive testing method, the acceptance criteria and any inspector certification are Not stated, and this page adds none. The site lists no defect label and no pass or fail label for this row.

Because inspecting comes before the cleanup, action segments matter: a spec that wants inspecting, grinding, cleaning and marking separated has to name them, and the site does not set an episode length. Whether an episode covers one weld, one part or a run of parts is a choice for the spec.

Only what the row says: "grind and clean seams". The site names an angle grinder and files on the two separate grinding rows of the same workshop, not on this row, so this page does not attribute a grinder, files or any cleaning tool to it. Grinder or tool brands, disc or wheel type, finish and cycle time are Not stated.

Contact events can mark when a hand meets the part or a tool, and action segments can mark the start and end of a grinding pass, where the spec asks for them, aligned to video and depth. Whether the cleanup is judged complete is Not stated: the site lists no completion label.

The row ends with "mark parts for rework". The marking tool, the mark itself and any rework decision rule are Not stated. The scribe, square and punch named on this site belong to the two separate layout rows of the same workshop, not to this row, so this page does not attribute them here.

Whether a program wants the marked part as the end of an episode, or a rework label beside it, is a choice for the spec. The site lists no rework or reject label for this row, and this page claims none.

The site says depth degrades on wet and reflective surfaces, and confidence maps mask the low-confidence regions. An episode is rejected when more than 35 percent of depth pixels in the manipulation region are invalid. The site does not say how depth behaves on bare, ground or freshly cleaned metal, so this page promises no figure.

Exposure handling near a work light or an inspection light is Not stated. Raise lighting in the spec and check it on a Spec Pilot.

A grinder, a guard and the part itself can sit between the lens and the fingers. The site states that fingers disappear behind objects, which is why each hand joint carries a visibility flag, and that reach and grasp frames are where blur and occlusion show up. The head camera is global-shutter, which avoids rolling-shutter skew.

The site gives no figure for how often a finger is hidden during grinding or marking, so a spec can ask for wrist views or a second angle and check it on a Spec Pilot. The site states no grip-force or tool-pose stream, and this page does not claim one.

The site makes no safety, protective equipment or compliance claim for any welding or fabrication environment, and this page makes none. PPE, sparks, fumes and machine guarding are Not stated. Welding codes, standards and certifications are Not stated.

For every delivered episode the site states that faces, plates and screens are blurred irreversibly, 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, an undocumented bystander is a stated reject criterion, and every episode links to a signed contributor release. The site states no separate rule for drawings, labels or machine displays in a shop, so this page does not claim they are blurred. Settle any of these in the spec before capture.

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 this row the site states 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 and tool occlusion is the first risk. A grinder, a guard or the part covers the fingers, which is why each hand joint carries a visibility flag. An episode is rejected when the manipulated object is out of frame for more than 15 percent of it. For this work the spec decides what that object is: the part, the seam or the grinder.

Invalid depth on reflective metal is the second, under the 35 percent rule above. 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 an undocumented bystander or for a mismatch against the named skill spec. A reject log ships with every batch.

Whether a weld was judged correctly, or a seam cleaned enough, is Not stated: the site lists no defect, pass, fail or completion label for this row. 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. The site names no weld inspection or metal fabrication 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. 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 speed applies to this row, so the timeline is quoted against your spec. Collections are billed per validated hour, and rejected episodes and recaptures are never billed.

Describe the steps and where each starts and ends: inspecting the weld, grinding and cleaning the seam, marking the part. Say whether an episode covers one weld, one part or a run of parts, what counts as the object and what counts as a completed step, 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.