GitHub - TDiblik/main-gate-alpr: Recognize license plates (and numbers) using fine-tuned yolov8, OCR (tesseract) and Hikvision camera
Sep 04, 2023 - github.com
The markdown data provides a detailed guide on how to set up and use the Main Gate ALPR (Automatic License Plate Recognition) system. It includes sections on how the system works, how to set it up, prerequisites, starting the websocket and matching server, starting the example client, training your own model, testing models, and configuring the environment. It also includes development notes and acknowledgements. The system uses a web server to load environment variables and AI models into memory. It then reads frames from an IP camera or video, passes them to YOLO for analysis, crops cars into an array, and checks the distance. It then passes the cropped image to a fine-tuned YOLO for license plates, crops and pre-processes the license plate, separates it into characters, passes it to Tesseract for reading, and finally validates and sends the license plate and cropped car to all websocket-connected clients. The system also includes options for saving results and inserting them into a database.
Key takeaways:
The Main Gate ALPR is a system that uses AI models to read license plates from a video feed. The system can be configured to save results, connect to a database, and send results to websocket-connected clients.
The system works by loading all environment variables and AI models into memory when the web server starts. It then reads frames from the video feed, passes them to the AI model for analysis, crops the images of cars and license plates, and processes the license plates to read the characters.
The system requires Python 3.11.2, Docker (optional), and Rust (optional). It also requires several AI models, which can be found in the `./ai/resources` folder. The system can be set up to start automatically using a systemd service.
The system can be configured using environment variables. These include settings for debugging, the web socket server port, the video input, the AI models to use, database settings, and various other options.
