Real Time Car Model and Plate Detection System by Using Deep Learning Architectures

The project presents an automated license plate detection and recognition system using deep learning architectures. The system detects vehicle license plates from uploaded images with YOLOv10, extracts alphanumeric text using Gemini 2.5 Flash API, and verifies the extracted plate number against a blocklist stored in a CSV file. If a match is found, an email notification with the processed image is sent via Gmail. The web application, developed using Flask framework, includes user registration, login, image classification, and logout modules. SQLite serves as the database for user management. The front-end is built with HTML, CSS, and JavaScript, ensuring a responsive and secure interface. YOLOv10 is trained on a public license plate dataset from Roboflow Universe, enabling accurate bounding box detection. The cropped plate region is sent to the Gemini API for text extraction, followed by blocklist comparison. The system supports image upload, processing, and result visualization. Testing shows reliable detection and extraction performance under varied image conditions. The solution integrates computer vision, natural language processing, and web development into a cohesive pipeline. Future improvements include video stream support and dynamic blocklist management.

Keywords: License Plate Detection, YOLOv10, Gemini 2.5 Flash, Blocklist Verification, Gmail Notification, Flask Framework, Deep Learning, Image Processing, Web Application, SQLite.