Dual-Perspective Household Task Videos

The goal is to build a small but neatly-structured video dataset that will feed a computer-vision model focused on everyday domestic activities. Each entry in the dataset must show a single household task—cleaning, cooking, or organizing—from two synchronized viewpoints:

• First-person: a head-mounted (egocentric) camera capturing what the performer sees.
• Third-person: a fixed, tripod-mounted camera giving a clear side or frontal view of the same action.

For every recorded task, both clips should start and end together, be labelled with the task name, and be delivered in lossless or high-quality H.264 MP4 at 1080p or better. Natural indoor lighting is fine as long as the objects, hands, and completed result are clearly visible.

Acceptance criteria
1. One paired recording per task instance (e.g., “wiping countertop,” “stowing pots,” “sorting bookshelf”).
2. Audio optional; if included, keep ambient household sounds only.
3. Camera shake on the head-mounted view should be minimal enough for frame-by-frame annotation.
4. Files named consistently: _FP.mp4 and _TP.mp4 plus a short text file describing room, objects, and approximate duration.

Feel free to use GoPro, Insta360, DSLR, or comparable gear—any model that can provide stable HD footage will work. Once the raw videos pass a quick quality check, I will confirm and release the milestone.