Capture catalogue · capture brief

Cleaning & Maintenance

This brief is scoped on demand. What that means, and what every brief includes.

What the brief covers

The brief specifies egocentric capture of household cleaning in residential homes in suburban Bangalore: wiping tables and surfaces, mopping and sweeping, folding laundry, and cleaning with tools. It mixes two kinds of manipulation: long, repeated strokes across a surface, and two-handed handling of a deformable object when laundry is folded. In tool-based cleaning, contact with the surface is made through a brush, a cloth or a mop rather than the hand. Both hands and the cloth or garment have to stay visible: a Clip where they leave the shot can be turned down as hands or task out of frame.

The specification

Set in the brief before anything is recorded.
Capture deviceRay-Ban Meta
Camera1920x1080 @ 30 FPS
Spatial tracksWide-angle camera lens intrinsics
AnnotationTier 1 · Tier 2
AvailabilityScoped on demand

What a delivery carries

Tiers 1 and 2 apply. Tier 1 carries each Clip's QA verdict, and the reasons a Clip can be turned down are a closed list of nine, published on /method-qa. Tier 2 gives each action segment its start and end frame, the hand that acts and a one-line description. Segmentation re-encodes rather than copying the stream, so each Clip starts on the measured boundary instead of the nearest keyframe. The lens intrinsics named in the specification are calibration data for the rig, not an FH-Ego v1 field, and travel with a delivery as the engagement sets. The Clips can be written as HDF5 beside their video, as LeRobot v2.1, or as FH-Ego v1 JSON with MP4.

Recording it

The work happens in lived-in homes, so other people may be in the room; any face the detector finds is blurred, and the delivered copy carries no sound. Glasses keep both hands free and the camera at eye level as the work moves from room to room.

What to know before you commission this

  • Ray-Ban Meta records 1920×1080 at 30 frames per second from a single camera, in clips of about three minutes exported through Meta's own app. Depth and wrist telemetry sit outside this rig.

Sample

Published sample: none

Other briefs in the catalogue

Commission this brief