Struct FaceProcessor 

pub struct FaceProcessor;

Face processing pipeline

Call process_photos to run the full detect -> embed -> cluster pipeline.

Implementations

impl FaceProcessor

pub fn run_clustering( face_repo: &FaceRepo<'_>, clustering_threshold: f32, resolver_weights: ResolverWeights, ) -> Result<usize, String>

Run incremental clustering in two stages:

1. Gallery k-NN retrieval + confidence bands: HIGH -> auto-assign to top cluster AMBIGUOUS -> queue for user review LOW -> leave for Stage 2
2. Complete-link agglomerative clustering on still-unresolved faces

Public so integration tests (and re-cluster-only flows that skip detection) can call it without going through the full process_photos pipeline.

pub fn regenerate_missing_crops(drive_path: &Path) -> Result<usize, String>

Regenerate missing face crop files from stored bounding box data.

pub fn faces_dir(drive_path: &Path) -> PathBuf

Get the face crops directory for a drive.

impl FaceProcessor

pub fn process_photos( drive_path: &Path, model_dir: &Path, detector_confidence: f32, clustering_threshold: f32, resolver_weights: ResolverWeights, progress_tx: Option<Sender<FaceProcessingProgress>>, cancel_flag: Option<Arc<AtomicBool>>, ) -> Result<FaceProcessingResult, String>

Run the full face processing pipeline on unprocessed photos.

Uses rayon for parallel detection/embedding with thread-local ONNX sessions, then batches DB writes for efficiency.