Stereo Surface Quality Inspection Data

It is first-person video of a person inspecting parts and items against a standard, recorded from their own point of view with two calibrated cameras and with hand pose and sensor streams alongside. The tasks are the ones the site’s catalog lists as inspection work: line quality inspection, food plant defect and foreign-object inspection, weld inspection, damaged-goods sorting, room checks and component soldering. 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 head-mounted phone or a multiview rig as the capture setup, 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. It records a person performing an inspection, not a library of defects. Firsthand is not affiliated with Foxglove or the MCAP maintainers.

The egocentric visual inspection page covers inspection as a task family: the look, handle, rotate, check and decide loop, the catalog rows, the ten-stream episode and the labels the site does not state. The data labeling and quality verification page covers how labels and batches are checked and rejected, which is a review process and not people inspecting parts. This page does not repeat either. It covers the two choices layered on top of a subset of the inspection rows: a calibrated stereo pair for depth, and MCAP as the delivery container.

The rows are the same and the claims are as limited. This page states no defect label, defect rate, detection accuracy, tolerance, measurement value or pass or fail label, because the site states none for these rows, and it states no safety or certification position for any of these environments.

Nine rows count, and every one lists a head-mounted phone or multiview rig as its capture setup. Two are quality inspection on a production line and In-Line Quality Inspection, in the Plastic & Light-Assembly Manufacturer environment. Two are food plant rows: inspecting and removing defective products under an inspection light, and QA and foreign-object inspection on the belt. Weld Inspection & Cleanup sits in the Metal Fabrication & Welding Workshop, Room Turnover & Inspection in Hotel & Lodging, two rows sort damaged goods in the back of a retail store, and Component soldering ends with inspecting the joints. The first table below reproduces each as the catalog states it.

No catalog row is named surface quality inspection, so this page does not say one is. The rows describe inspecting units, items, welds and a room against a standard. Which of them involves a surface, and what counts as a defect on it, is set in the spec. Eight rows are Ready to deploy and the soldering row is Available on request.

Inspection is about what a person sees and how they change the view to see it. A stereo pair recovers depth from the calibrated frame pair rather than from a standalone depth sensor, so the form of the part and the station sits in the same frame as the hands that turn it. The site gives no depth accuracy figure and no defect label, so whether a given scratch, seam or edge is resolvable in depth is Not stated.

The hard-environments page states that reflective metal, small parts, clutter and awkward lighting are where a single camera loses depth. It does not say that stereo resolves a surface defect, and this page does not either. Which station, part and standard a collection covers is set by the row and agreed in the spec.

The site describes two stereo setups and does not say which one an inspection order 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.

Which rig, resolution and streams an inspection order uses is Not stated on the site, and the second table below sets what is stated beside what is not. The site does not state a camera count, frame rate or lighting setup specific to inspection either, including how the inspection light named in one catalog task is lit or measured. Intrinsics and extrinsics are re-solved at the start and end of every session, and streams are hardware-triggered onto one clock with a 2.0 ms cross-stream ceiling.

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’s 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 inspection work the site states head and wrist RGB, depth, hand pose, contact and grasp events and action segments in its episode schema, plus IMU on the stereo rigs. Instance segmentation is stated for warehouse collections and not for inspection. What the site does not state is which of these 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 third table below sets each stream beside what the site states about it.

The part is the first risk. An inspector steadies it with one hand and turns it with the other, and fingers can hide the very face being checked. The visibility flag on each joint is the answer the site gives, contact events mark when a hand meets the part, and a spec can ask for wrist views so close-range detail is still recorded.

Depth is the second. The hard-environments page names glossy and dark surfaces as a source of lost depth, and the site says depth degrades on wet and reflective surfaces and masks low-confidence regions with per-frame confidence maps. 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 a specific part or finish, so this page promises no figure.

Calibration is the third, because stereo depth is only as good as the rectification behind it: calibration is 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. At an inspection station, coworkers, monitors and paperwork can enter the frame, which is why faces and screens are blurred irreversibly at ingest and a reviewer confirms the result.

Defect labels, a defect taxonomy, a pass or fail outcome label for each inspected item, measurement values such as dimensions or readings, and part, serial or lot identity are not stated on the site as standard labels. Some catalog task names mention spotting defects, measuring critical dimensions or rejecting failures, but those describe what the operator does, not a label delivered with the episode.

This page gives no defect rate, detection accuracy or tolerance, and stereo capture does not add one. If your policy needs any label of this kind, it is defined in the written spec as annotation depth and covered in the quote before capture begins.

Off-the-shelf covers what is already in stock: the stereo category lists the stereo datasets and hand-tracking video, and only dataset 06 is stated as MCAP. This page does not claim off-the-shelf coverage of inspection 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.

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 catalog status on each task row tells you which case you are in. Collections are billed per validated hour.

Describe the item, the station, the standard it is checked against, the decision at the end and what counts as one completed check, plus 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 fourth table below shows what each spec element decides.