Car tracking & Number plate recognition

Developed by Genesys Research Lab, this project is a comprehensive computer vision solution designed to automate traffic monitoring and secure access control. The system replaces manual vehicle logging with an intelligent pipeline capable of processing live video streams to detect, track, and identify vehicles with high precision. Key Features: Advanced Detection Pipeline: Utilizes YOLOv11x for detecting vehicles (Cars, Trucks, Buses) and a custom-trained YOLOv11n model (trained on 8,000+ frames) specifically for locating license plates. Robust Tracking: Implements DeepSORT algorithms to assign unique IDs to vehicles, ensuring consistent tracking across frames and preventing double-counting even in occluded scenes. Optical Character Recognition (OCR): Integrates PaddlePaddle OCR to extract alphanumeric text from license plates, storing the data instantly in a structured database. Customizable Monitoring: ROI Selection: Users can draw a Region of Interest to focus processing power only on relevant road areas. Virtual Traffic Lines: Configurable "IN" (Green) and "OUT" (Red) lines allowing the system to determine travel direction and automate entry/exit timestamps. Data Management: A user-friendly dashboard displays live detections with confidence scores. It includes a searchable historical database where users can filter logs by Track ID, number plate, or date, and export reports via CSV. Technical Stack: Models: YOLOv11x, YOLOv11n, DeepSORT, PaddleOCR Input: RTSP / CCTV Streams, Video Files Hardware Optimization: Optimized for GPU acceleration (Benchmarked on RTX 2080 Ti). Use Cases: Ideal for smart parking management, toll booth automation, gated community security, and traffic flow analysis.