• Built a real-time vehicle license plate detection and recognition system using YOLOv8 for detection and
PaddleOCR for text extraction.
• Achieved 98.8% detection accuracy (mAP@50) on 24k+ annotated frames with reliable end-to-end OCR
performance.
• Deployed as an interactive Streamlit app supporting live video input, automatic cropping, logging, and
confidence filtering.
• Tools Used: Python, YOLOv8 (Ultralytics), PaddleOCR, Streamlit, OpenCV