Extraction and Classification of Bianxiu Patterns Based on RepVGG

This study proposes a novel method for the extraction and classification of Bianxiu patterns based on the RepVGG deep learning model. Bianxiu, a form of traditional Chinese embroidery, is renowned for its intricate designs and deep cultural significance. The complexity of its patterns makes it a challenging task for automatic recognition and classification. By leveraging the RepVGG model, known for its efficient architecture and superior performance in computer vision tasks, this paper introduces several innovations to enhance the pattern recognition process. Key contributions include: 1) the development of an efficient feature extraction approach using RepVGG, which effectively captures fine-grained details while ensuring computational efficiency; 2) the integration of multi-level feature fusion to strengthen the models performance in fine-grained classification tasks; and 3) the implementation of a multi-task learning framework based on semantic embedding, which enables the simultaneous optimization of pattern classification and semantic interpretation. This work provides a solid foundation for advancing the automation of traditional embroidery pattern recognition and classification.