CRNN-Based Abstract Artistic Text Recognition

As an important technique for visual perception, text recognition has been an active and long-standing research topic in the field of computer vision. With the rapid development of deep learning in recent years, significant progress has been made in scene text recognition, leading to the proposal of numerous excellent algorithms. However, limited research has been conducted on the recognition of abstract artistic text, which possess unique application scenarios. In this work, we improved and optimized various aspects of the classical Convolutional Recurrent Neural Network (CRNN), including data training, data preprocessing, and feature extraction. While ensuring the delivery of model’s ability in common text recognition tasks, we have greatly enhanced its end-to-end capability to spot abstract artistic text of arbitrary length. Additionally, we performed extensive experiments on a large number of artistic text recognition, comparing our model with the classical CRNN model and Baidu’s Paddle OCR tool. The results demonstrate the significant advantages of our model in this regard.