Automating Driving Tests with Computer Vision

The Automated Driving Test System (ADTS) was developed for the Gujarat Motor Vehicle Department to replace manual driving assessments with real-time, camera-based evaluation.

Problem

Traditional driving tests depend heavily on manual supervision, which can be inconsistent and time-consuming. The goal was to automate the process using multi-camera setups and backend systems capable of evaluating candidates objectively.

Approach

We designed a system that integrates:

• Multi-camera RTSP streams for real-time visual analysis
• Computer vision pipelines using OpenCV for lane tracking, stop-line detection, and rule violations
• Python FastAPI backend to handle data flow, candidate information, and result computation
• PostgreSQL for persistent data storage and test results

Results

• Fully automated evaluation with zero manual intervention
• Real-time scoring aligned with official driving parameters
• Seamless integration with government hardware and dashboards

The project demonstrated how AI + backend engineering can eliminate human bias and improve large-scale public systems.