Controlo Acessos @ Convento Cristo
Computer vision system for counting and tracking people in the Charola of the Convent of Christ using YOLO and DeepSort.
Overview
The Controlo de Acessos @ Convento Cristo project consists of a computer vision system to count and track people entering and exiting the Charola of the Convent of Christ in Tomar. The system uses real-time video feeds to support visitor management at a UNESCO World Heritage monument.
Approach
The system combines YOLO for real-time object detection with multi-object tracking algorithms such as DeepSort. This pipeline detects people in each video frame, tracks individuals across frames, and counts entries and exits by monitoring trajectories across defined boundary lines.
Background
Gonçalo Sousa initiated the project as his MSc thesis, integrating YOLOv5 and DeepSORT into an early research prototype and building a Flask-based web dashboard for visitor reporting. His work validated the feasibility of ML-based visitor counting at the site.
Sofia Vendor is now continuing the project as her BSc final project. Her focus is on revamping the solution with newer YOLO models, refining the tracking pipeline, and taking the system towards an in-situ prototype for real-conditions testing at the Convent of Christ.