HIFA-LPR: High-Frequency Augmented License Plate Recognition in Low-Quality Legacy Conditions via Gradual End-to-End Learning

Scene text detection and recognition, such as automatic license plate is a technology utilized in various applications. Although numerous studies have been conducted to improve recognition accuracy, accuracy decreases when low-quality legacy images are input into module due low image quality lack of resolution. To obtain better this study proposes high-frequency augmented model which the super-resolution integrated trained collaboratively via proposed gradual end-to-end learning-based optimization. optimally train our model, we propose holistic feature extraction method that effectively prevents generating grid patterns from super-resolved during training process. Moreover, exploit information affects performance based on augmentation. Furthermore, learning process weight freezing with three steps. Our three-step methodological approach can properly optimize each provide robust performance. The experimental results show superior existing approaches conditions UFPR Greek vehicle datasets.