Proposing Vietnamese Text Recognition Algorithm Combining CRAFT and VietOCR

Abstract

Letter recognition based on artificial neural networks is currently achieving promising results. In this paper, we propose a method that combines character-region awareness for text detection (CRAFT) and VietOCR to recognize letters and Vietnamese texts effectively based on finding characters in words and affinity among them. The algorithm uses the tool to crop each character to put it into VietOCR to recognize letters. The results show that the algorithm achieves an accuracy of 89.84% with 10 epochs. We find that the proposed method has accuracy and execution time depending on the variety of input data based on the obtained results.