About this project
This final-year project (projet de fin d'études) implements a robust traffic sign recognition system using a two-stage deep learning pipeline. YOLOv8 processes real-time video streams to detect and localize traffic signs, then a CNN classifier further categorizes each detected sign into one of 43 specific categories.
The system handles both live camera feeds and pre-recorded video, making it suitable for driver assistance systems, autonomous vehicle pipelines, and traffic management applications. The two-stage approach separates detection from classification, allowing each model to be optimized independently.
Key features
YOLOv8 detection stage
Processes real-time video to detect and localize traffic signs with bounding boxes.
CNN classification stage
Classifies each detected sign into one of 43 specific categories.
Real-time & video support
Works on live camera feeds and pre-recorded video files.
ADAS-ready architecture
Suitable for driver assistance systems and autonomous vehicle pipelines.