Implementation of OCR system on extracting information from Vietnamese book cover images.
Tham Nguyen Thi, Trong-Hop Do, Myungsik Yoo · 2023
This paper describes an OCR system for extracting information on Vietnamese book cover images. The set system takes as input the book cover image and outputs the book title, author, publisher and other information. The system is composed of three modules including text detection, text recognition and information synthesis. For the text detection module, we experiment with EAST and SAST models. For the text recognition module, we experimented with CRNN, SVTR, Transformer OCR models. In the last step, we use Yolov4 for object detection then to process the information. Our system extracts book cover information with accuracy per word up to 84.06% and on WER and CER score respectively 22.67% and 22.41%.