Campus Entry/Exit AI Control

Real-time person/vehicle counting, ID tracking and reporting from multiple camera streams at Erciyes University gates and tram stops.

ERÜ BİDB — Internship Project (1 Month) In Production — Used by the Rectorate
Campus gate — real-time person/vehicle detection

Project Information

  • Institution: Erciyes University — Department of Information Technology
  • Scope: Gates & tram stops; counting of people, bicycles, motorcycles
  • Features: Real-time counting, ID tracking, multi-camera support, reporting
  • Status: Live usage (In Production)
  • Duration: 1 month (internship period)

Overview

The system takes live streams from the university's existing camera infrastructure to detect entities such as people, bicycles and motorcycles, assigns an ID to each, and reports entry-exit statistics in real time. By combining YOLO-based detection with an extensive computer vision system, an end-to-end solution focused on speed and efficiency was designed. The system is already in use by the Erciyes University Rectorate.

End-to-End Setup: In addition to the feasibility, data access, model development, testing and go-live stages, the procurement and installation of high-compute GPU/Server resources were also included in the project scope.

Architecture & Components

  • Data Ingestion: Pulling video streams from the existing camera infrastructure.
  • Preprocessing: Frame normalization, noise reduction, ROI enhancement, lighting/contrast adjustments (OpenCV).
  • Detection: Person/bicycle/motorcycle detection with the YOLO family; optimizations suited to real-time operation.
  • ID Tracking: Identification of entities in a multi-camera scenario and tracking them within the stream (ID continuity).
  • Counting & Logic: Virtual line/zone-based entry counts, density analysis, timestamped logging.
  • Reporting: Real-time dashboard, historical statistics, hourly/daily/weekly summaries; data export.

Technology Stack

Python PyTorch YOLO OpenCV CUDA Django SQLite Linux

Deployment & Operations

  • Server: GPU-accelerated server setup and configuration.
  • Scaling: Real-time monitoring & counting across 10+ camera streams.
  • Privacy: Generation of statistics containing no personal data; logging only at the counting and ID level.
Go to Projects