ALDEN BOBY

Number Plate Detection in Different Environments using Generative Adversarial Networks

Supervisor: Dr. Dane Brown

In real-world licence plate recognition scenarios, images are not as straightforward as the 'toy' datasets used to test the performance of existing systems. Real-world data is typically 'dirty' as it may contain poor lighting and occlusion, obscuring the information on a licence plate, making it difficult to detect. Cleaning input data before using it for licence plate recognition is a complex problem, and existing literature addressing the issue is still limited.

This research uses two deep learning techniques to improve licence plate visibility towards more accurate licence plate recognition.  A one-stage object detector called You Only Look Once (YOLO) is implemented for locating licence plates under challenging situations. Super-resolution Generative Adversarial Networks (SRGANs) are considered for image upscaling and reconstruction to improve the clarity of low-quality input. 

The main focus involves training these systems with datasets that include difficult to detect licence plates, enabling better performance in unfavourable conditions and environments, leading to increased accuracy.

Github - https://github.com/AldenBoby/LPR-Using-GANs