Crate capture_utils 

Filesystem-based recorder implementations and dataset management utilities.

This crate provides:

• JsonRecorder: Default recorder that writes frame metadata to JSON files.
• Overlay generation from label JSON files (draws bounding boxes on images).
• Dataset pruning utilities for filtering and copying runs.

Integrates with vision_core::interfaces::Recorder trait and data_contracts schemas.

Structs

JsonRecorder

Default file-based recorder: writes frame metadata/labels to run_dir/labels/frame_XXXXX.json.

Functions

build_capture_metadata

build_inference_metadata

Helper for inference outputs: callers set label provenance on Label before recording.

generate_overlays

Generate overlay PNGs from label JSONs in a run directory.

prune_run

Prune a run directory into a destination root, copying kept artifacts.