7a99ui/AutoPlateTN

A complete end-to-end system for detecting vehicles, localizing license plates, enhancing plate images, and performing OCR to extract Tunisian plate numbers. This project combines object detection, image processing, and deep-learning OCR to build a full ALPR pipeline suitable for real-time or batch processing. # AutoPlateTN - Automatic Tunisian License Plate Recognition System An end-to-end deep learning pipeline for detecting and recognizing Tunisian license plates in images. Built with state-of-the-art object detection and OCR models, featuring an interactive web interface. Demo • Features • Quick Start • Results • Usage --- ## Demo ### Complete Pipeline Visualization #### 1️ Vehicle Detection *Cascade R-CNN detects all vehicles in the image with bounding boxes* #### 2️ License Plate Detection *Faster R-CNN localizes license plates on detected vehicles* #### 3️ Plate Enhancement *Advanced image preprocessing: denoising, upscaling (4x-6x), CLAHE, binarization* #### 4️ Final OCR Results *CRNN model reads characters and formats plates in Tunisian standard (Arabic + digits)* --- ## Features ### **Multi-Stage Detection Pipeline** - **Stage 1**: Vehicle Detection using Cascade R-CNN - **Stage 2**: License Plate Detection using Faster R-CNN - **Stage 3**: Adaptive Image Enhancement (4 methods) - **Stage 4**: OCR Recognition using CRNN with CTC Loss ### **Advanced Image Processing** - **4 Enhancement Methods**: Original, Light, Full (recommended), Aggressive - **Techniques**: - Denoising (fastNlMeans) - Super-resolution upscaling (4x-6x) - Contrast enhancement (CLAHE) - Gamma correction - Bilateral filtering - Adaptive binarization - Morphological operations ### **Interactive Web Interface** - Built with **Streamlit** for easy deployment - Real-time processing with live visualization - Adjustable confidence thresholds - Multiple enhancement method selection - Downloadable results ### **Tunisian Plate Support** - Recognizes both standard and private plates - Automatic formatting with Arabic text (تونس / نت) - Proper number grouping and centering - Format: `XXX تونس XXXX` or `XXXXXX نت` --- ## Quick Start ### Prerequisites - Python 3.8+ - CUDA-capable GPU (optional, for faster processing) - Git ### Installation 1. **Clone the repository** ```bash …