Computer Vision Project

QPlate Vision

A CNN-based OCR system developed to detect and recognize Qatari license plates using YOLOv5m, OpenCV preprocessing, and EasyOCR.

Python YOLOv5m OpenCV EasyOCR OCR Computer Vision

Project Overview

QPlate Vision is a localized automatic license plate recognition system designed for Qatari vehicles. The project focuses on detecting vehicles, isolating relevant regions, enhancing image quality, and extracting plate numbers under real-world conditions such as lighting variation, blur, and angle changes.

My Contribution

Detection & Testing

Contributed to the YOLOv5 integration, testing workflow, and debugging process across the recognition pipeline.

Ground Truth Dataset

Worked on creating the ground truth dataset used to compare OCR outputs against expected plate values during evaluation.

How It Works

Vehicle detection using YOLOv5m

Bounding-box based region isolation

Image preprocessing with grayscale, blur, and CLAHE

Text extraction using EasyOCR

Ground truth comparison and evaluation

Performance measurement with precision, recall, F1-score, and accuracy

Performance Snapshot

95%

Precision

92.68%

Recall

93.83%

F1-Score

88.37%

Accuracy

Project Gallery

Example of detected vehicle with bounding box

Example of preprocessed cropped vehicle image

Why This Project Matters

This project demonstrates how computer vision and OCR can be adapted to localized, real-world challenges. By focusing on Qatari license plates, the system addresses bilingual formatting, varied environmental conditions, and practical recognition needs relevant to traffic, parking, and smart-city applications.