License plate detection using a deep transfer learning approach

Abstract

License plate detection in Bangladesh poses unique challenges due to the diverse designs, fonts, and variations in plate structures. This study provides a comprehensive overview of recent advancements in License plate detection, introducing an innovative approach utilizing deep learning models. Several integrated models for image classification were evaluated in the research, including MobileNetV2, VGG-16, DenseNet201, ResNet-50, EfficientNet-b0 and Inception-V3. The findings of the study revealed that MobileNetV2 emerged as the top performer, achieving the highest accuracy of 96%. This model demonstrated exceptional performance in accurately identifying License plates. On the other end of the spectrum, EfficientNet-b0 exhibited the lowest accuracy at 90%. The remaining models, including VGG-16, DenseNet201, ResNet-50 and Inception-V3 showcased varying degrees of accuracy falling between these two extremes. This research underscores the potential of transfer learning models, in enhancing disease detection accuracy. It emphasizes the significance of continuous innovation and improvement in License plate detection methodologies to support sustainable and cost-effective processes.