okere-rafiatou/trafficvision-group10

Real-time road traffic detection, tracking and counting using YOLOv11 + ByteTrack AIMS Senegal CV Project, Group 10 --- title: TrafficVision colorFrom: yellow colorTo: red sdk: docker pinned: false app_port: 7860 --- # TrafficVision — Real-Time Road Traffic Detection and Tracking **Group 10** | AIMS Senegal | Computer Vision Project 2025–2026 Members: Moute Jean-Baptiste, Okere Rafiatou, Gaolatlhe Angelah Kgato Supervisor: Dr. Jordan Felicien Masakuna --- ## Overview TrafficVision is a real-time computer vision system for detecting, tracking, and counting road-traffic objects across multiple video scenes. It uses a fine-tuned YOLOv11 model combined with the ByteTrack multi-object tracking algorithm to assign persistent identities to detected objects across frames, enabling accurate unique-object counting rather than simple per-frame detection counts. The system is part of a collaborative project at AIMS Senegal where each group contributes traffic data in a shared format to build a global multi-scene traffic monitoring dataset and dashboard. --- ## Features - **Object detection** using fine-tuned YOLOv11n across six traffic classes: person, bicycle, car, motorcycle, bus, truck - **Multi-object tracking** with ByteTrack — each object receives a unique, persistent ID across frames - **Unique counting** — each object is counted only once per session, regardless of how many frames it appears in - **Speed estimation** in pixels per second from centroid displacement between frames - **Direction detection** — up or down movement relative to the frame - **Line-crossing events** — detects when an object crosses the horizontal counting line at mid-frame - **Three video input modes** — local file upload, YouTube URL (direct stream), or webcam - **Live annotated stream** — real-time bounding boxes, class labels, and track IDs displayed in the browser - **19-column CSV logs** following the shared schema for global dashboard integration - **Analytics dashboard** — per-scene object counts, class distribution charts, traffic intensity timeline, and dataset download --- ## Project Stru …