yscv-track

Multi-object tracking: DeepSORT, ByteTrack, Kalman filter, Hungarian assignment, and Re-ID.

use yscv_track::*;

let mut tracker = Tracker::new(TrackerConfig::default());
loop {
    let detections = detect_objects(&frame);
    let tracked = tracker.update(&detections);
    for t in &tracked {
        println!("Track {} at ({}, {})", t.track_id, t.x, t.y);
    }
}

Algorithms

Re-ID

• Color histogram features (no model needed)
• Cosine similarity matching
• VP-Tree approximate nearest neighbor
• Feature gallery with FIFO eviction (max 100 per track)

Tests

57 tests covering track lifecycle, assignment, Kalman prediction, Re-ID matching.