Real-time Classifier of Multilingual Font Styles based on ResNet, SwordNet, Logistic Regression and Random Forest Algorithms

Abstract

Different languages have different characters. At the same time, each character has a lot of font styles. This makes it difficult for humans to recognize different font styles for different characters. However, being able to detect and identify these font styles quickly and accurately has many important application use cases in different fields. At the same time, a large number of Internet users use web pages to query font styles. Therefore, I choose to make this real-time multilingual font style recognition algorithm. In this paper, I propose an algorithm that recognizes the input text and pictures in real time to judge the language and style of the text. It includes ResNet, SwordNet, logistic regression and random forest algorithms. The whole algorithm also calls pytesseract and Google Tesseract to realize text recognition and text positioning. I used Font Datasets used in "Font and Calligraphy Style Recognition Using Complex Wavelet Transform" for training. At the same time, I also built an image text recognition algorithm and generated various font styles as a data source. Based on this data, we adjusted the parameters and finally achieved an accuracy rate higher than 90%.