Crate yscv_recognize 

yscv-recognize

Face and object recognition with embedding matching and VP-Tree approximate nearest neighbor search.

use yscv_recognize::*;

let mut recognizer = Recognizer::new();
recognizer.enroll("alice", &embedding_alice);
recognizer.enroll("bob", &embedding_bob);

let matches = recognizer.identify(&query_embedding, 0.7);
// matches[0].label == "alice", matches[0].score == 0.95

Features

• Cosine similarity matching between embeddings
• VP-Tree (Vantage Point Tree) for fast approximate nearest neighbor
• Gallery management: enroll, remove, snapshot, restore
• Serializable: save/load gallery as JSON

Tests

16 tests covering enrollment, matching, gallery persistence.

Modules

vp_tree

Structs

IdentityEmbedding

IdentitySnapshot

Recognition

Recognizer

RecognizerSnapshot

VpTree

Vantage-point tree for approximate nearest-neighbor search.

Enums

RecognizeError

Constants

CRATE_ID

Functions

cosine_similarity

cosine_similarity_slice