annie.media.compose

Compose a video’s vdet and track annotations into per-frame overlays (service).

A VideoEntry aggregates one video’s raw detections and all of its tracks. To draw a frame we merge the relevant boxes into a single FrameAnnotation, then hand it to annie.color.draw_overlay(), which colours each box by the standing rules:

• vdet boxes (track_id is None) → flat blue;

• track boxes → a stable unique colour per track id (never blue/green);

• the active / protagonist track → green, overriding its palette colour.

Two inclusion modes are used: the Browse five-frame strip shows vdet + only the active track; the full render shows vdet + every track.

Functions

load_entry_annotations(→ tuple[dict[int, ...)	Load and index a video's vdet and track annotations by frame.
merge_frame(→ annie.core.models.FrameAnnotation)	Combine vdet and selected track boxes for one frame.
strip_track_ids(→ list[int])	Track ids to draw on the Browse strip: just the active one, if any.

Module Contents

annie.media.compose.load_entry_annotations(entry: annie.core.models.VideoEntry) → tuple[dict[int, annie.core.models.FrameAnnotation], dict[int, dict[int, annie.core.models.FrameAnnotation]]]

Load and index a video’s vdet and track annotations by frame.

Parameters:

entry – The video whose annotation files to read.

Returns:

A (vdet_by_frame, tracks_by_id) pair where vdet_by_frame maps a frame index to its raw detections, and tracks_by_id maps each track id to its own frame_index -> FrameAnnotation mapping.

annie.media.compose.merge_frame(frame_idx: int, vdet_by_frame: dict[int, annie.core.models.FrameAnnotation], tracks_by_id: dict[int, dict[int, annie.core.models.FrameAnnotation]], include_track_ids: list[int]) → annie.core.models.FrameAnnotation

Combine vdet and selected track boxes for one frame.

Parameters:

• frame_idx – The frame to assemble.

• vdet_by_frame – Per-frame raw detections (from load_entry_annotations()).

• tracks_by_id – Per-track per-frame annotations (from load_entry_annotations()).

• include_track_ids – Which track ids to include on this frame.

Returns:

A single FrameAnnotation holding the vdet boxes followed by the selected track boxes present on that frame.

annie.media.compose.strip_track_ids(entry: annie.core.models.VideoEntry) → list[int]

Track ids to draw on the Browse strip: just the active one, if any.

Parameters:

entry – The video being shown.

Returns:

[active_track_id] when a protagonist is assigned, else [].