annie.parsers.vdet

Parser for .vdet files (raw per-video face detections).

A .vdet is computed once per video and holds all face detections, frame by frame. A single frame may contain several rows (several faces), so the loader groups rows by frame_id. Detections carry no track identity.

Functions

load_vdet(→ list[annie.core.models.FrameAnnotation])

Load a .vdet file into per-frame annotations.

Module Contents

annie.parsers.vdet.load_vdet(path: str | pathlib.Path) → list[annie.core.models.FrameAnnotation]

Load a .vdet file into per-frame annotations.

Rows sharing a frame_id are grouped, so a frame with three detected faces yields one FrameAnnotation holding three boxes. Boxes have track_id is None because raw detections are not yet tracked.

Parameters:

path – Path to the .vdet file.

Returns:

Per-frame annotations ordered by ascending frame index.