11 / 11 · ROAD
Perception for autonomous driving
A visual recognition system for autonomous driving: pedestrians, distance and lane markings in real time, in bad weather.
The problem
Perception that works on a sunny test video fails at night, in fog, snow or rain, or with the sun low on the horizon.
What I did
A GPU-accelerated YOLO model for pedestrians with dynamic distance estimation, a Canny and Hough lane pipeline tuned for adverse conditions, and an NLP voice assistant with geolocation.
What changed
Real-time detection that holds up in low light, snow and fog. It is the computer vision half of how I work.