SMFNet: One-Shot Recognition of Chinese Character Font Based on Siamese Metric Model

This paper focuses on Chinese character font recognition and introduces a novel one-shot learning model SMFNet for Chinese character font recognition, by thoroughly analyzing the fundamental characteristics of Chinese characters. Due to the considerable quantity and intricate structures of Chinese characters, conventional methods for font recognition face challenges in meeting the necessary requirements. To address this issue, a Siamese metric structure is incorporated as the model framework. This is complemented by coordinate attention modules and multiple loss functions, enabling one-shot font recognition across various categories such as standard printed fonts, soft brush calligraphy fonts, and hard pen calligraphy fonts. The model is trained and tested using the XIKE-CFS Chinese character font style dataset. Experimental results indicate that SMFNet achieves a recognition accuracy of 97.50% on the XIKE-CFS dataset and supports one-shot recognition with an accuracy rate of 92.84%. Furthermore, compared to other prevalent CNN classification models, SMFNet exhibits outstanding performance in terms of parameter quantity and demonstrates excellent capabilities in recognizing new fonts. In summary, this work provides innovative perspectives and methodologies for addressing challenges in the field of Chinese character font recognition.