An improved convolutional block attention module for Chinese character recognition

Abstract

Recognizing Chinese characters in natural images is a very challenging task, because they usually appear with artistic fonts, different styles, various lighting and occlusion conditions. This paper proposes a novel method named ICBAM (Improved Convolutional Block Attention Module) for Chinese character recognition in the wild. We present the concept of attention disturbance and combine it with CBAM (Convolutional Block Attention Module), which improve the generalization performance of the network and effectively avoid over-fitting. ICBAM is easy to train and deploy due to the ingenious design. Besides, it is worth mentioning that this module does not have any trainable parameters. Experiments conducted on the ICDAR 2019 ReCTS competition dataset demonstrate that our approach significantly outperforms the state-of-the-art techniques. In addition, we also verify the generalization performance of our method on the CTW dataset.