Stereo Vacuuming and Surface Sanitizing Data

It is first-person video of a person doing routine commercial cleaning, recorded with two calibrated cameras, with hand pose and sensor streams alongside. People reach it through several phrasings: vacuuming robot training data, stereo surface sanitizing video, an egocentric janitorial cleaning dataset, or custom commercial cleaning data collection at scale. Stereo describes the capture, and MCAP describes the container. They are separate choices.

Two, listed in lib/commercial-catalog-data.ts under the Commercial Cleaning & Janitorial environment, in the Facilities & Cleaning sector. "Vacuuming" reads "Vacuum carpets and hard floors in offices and corridors, moving chairs as needed" and is rated Easy. "Surface sanitizing" reads "Spray and wipe desks, handles and high-touch surfaces with sanitizer" and is rated Easy. Both are Active collection and both list "Phone (head-mounted) / multiview rig" as the capture setup. The environment lists offices, shopping malls, airports and industrial facilities as examples. The site says nothing about how many hours exist for either row, so this page claims no delivered hours. The first table below reproduces the rows.

This page is the stereo and MCAP treatment of two Easy rows together. The egocentric vacuuming, wiping and surface cleaning and commercial janitorial pages are first-person video that is not stereo as its subject. The stereo floor mopping page has mopping as its subject and does not name Surface sanitizing; the stereo carpet extraction and restroom pages cover different rows. 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. The stereo head rig the site describes has dual fisheye cameras at 1080p per eye on a 62 mm baseline, per-unit calibration, a 9-axis IMU and hand-tracking output, shipped as MCAP. Camera count, resolution and frame rate for cleaning work are Not stated.

The Vacuuming row names two floor types, carpets and hard floors, two places, offices and corridors, and one extra step: moving chairs as needed. A spec decides whether moving a chair is part of the episode or cut out, and where a pass starts and ends. In stereo, the floor, the chair and the hands on the tool are recorded in the same frame pair, and a wrist-mounted camera is available alongside the head camera for a closer view of the grip.

How depth behaves over carpet versus a hard floor is Not stated, so confirm it on a Spec Pilot. Vacuum brands or models and cleanliness or performance claims are Not stated and are not claimed here. Whether transit between rooms is kept or cut is set in the spec.

The Surface sanitizing row names a spray step, a wipe step and a set of targets: desks, handles and high-touch surfaces. These are short two-handed sequences on objects of very different shapes, and a calibrated pair records how the hands, the spray tool, the cloth and the surface sit relative to each other. Action segments can separate spraying and wiping when the spec names them, and contact events mark when a hand meets an item.

The site states that wet and reflective surfaces degrade active-stereo depth and that per-frame confidence maps are shipped so you can mask rather than guess. How a stereo pair behaves on a freshly sprayed surface is Not stated. The site states no sanitizer product, dwell time or cleaning standard for this row, and this page gives no chemical-handling, disinfection-efficacy or equipment-safety guidance.

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.

Desks hold screens, papers and belongings. Whether signage, documents or personal items are blurred beyond that general rule is Not stated, so it is settled in the spec. This is not legal advice, and the executed agreement controls.

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. From the site’s own statements, a file for these rows can carry the stereo pair, hand pose and IMU. Topic names, schemas and rates are Not stated and are agreed in the skill spec.

Hand occlusion is the first risk: a hand, a handle, a cloth or a moved chair can hide the fingers, which is why hand pose carries a per-joint visibility flag. Tool and surface changes are the second, since the same person moves from a vacuum to a spray tool to a cloth, and where each step starts is set in the spec. Calibration is measured at the start and end of every session under a 0.28 px RMS gate.

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 an undocumented bystander or for a mismatch against the named skill spec. A reject log ships with every batch. How often these failures occur on cleaning work is Not stated.

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 household chores dataset page lists vacuuming with obstacle negotiation in whole homes, but none of the datasets names office, corridor or high-touch surface work in a commercial building. For stereo capture of these two rows 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, so the timeline is quoted against your spec. Collections are billed per validated hour, and rejected episodes and recaptures are never billed.

Most programs begin with a Spec Pilot, a small first batch that proves the spec and accept criteria on real footage before you scale.