Autonomous Driving for Egyptian Streets
Developed a computer vision system for autonomous driving in Egyptian street conditions using 5,000+ annotated images and six models for lane detection, obstacle avoidance and traffic sign recognition, reaching 92%+ detection accuracy with real-time decision-making, extended with V2V and V2I communication.
Egyptian roads break most off-the-shelf driving models: faded lanes, unpredictable traffic, irregular signage.
The stack combines classical OpenCV lane extraction with trained detectors for obstacles and local traffic signs, fused into a decision layer that runs in real time on vehicle hardware. V2V and V2I messaging lets vehicles share hazard and intersection state, extending perception beyond line of sight.
Tools: OpenCV, TensorFlow, PyTorch, computer vision, V2X communication.
