Automatic inventory system of librarian books based on a deep learning algorithm with EAST and CRNN

Abstract

Although the researchers have made great progress in the field of text detection and text recognition, text detection and text recognition are still facing great challenges because of the differences of text fonts and the complexity of backgrounds. Traditional text detection and recognition methods rely on artificial designed features and rules, thus the methods usually requires higher text layout and text resolution. Aiming at the problem of automatic book inventory in library, the paper proposes a new method based on an EAST (Efficient and Accurate Scene Text Detector) detection and an CRNN (Continuous Recurrent Neural Network) recognition. In this method, the library book titles are detected by EAST to get the text area on the side of the books and also to output coordinates. Then, the content of the text area is further identified by the CRNN. Finally, through the comparison of the database, we know whether books of libraries are on corresponding shelves or not. The experimental results show that this method can quickly and accurately realize the task of automatic book title recognitions, and it can still effectively detect the text area and accurately recognize the book title in the case of dark light. Therefore, the method effectively solves the problem of manual inventory of books in existing libraries, which is time-consuming and laborious, and has a certain engineering application prospect.