Decorative Pattern Segmentation Using an Improved DeepLabv3 + Model

ABSTRACT This paper addresses the issue of neglecting the cultural connotations of traditional decorative patterns in image segmentation research. It proposes an image segmentation method for decorative patterns based on the improved DeepLabv3+ model. Images of decorative patterns and symbolic features in the Xiangtang Grottoes are acquired and preprocessed. Extension semantics is used to categorize the decorative patterns and extract semantic features of cultural symbols using representation models. The improved DeepLabv3+ model is used to segment the decorative pattern images. A comparison of the proposed model with classical models indicates the former's superiority in segmentation speed, segmentation performance, and the number of parameters. This method has high segmentation accuracy for extracting decorative patterns from images. This research provides a foundational approach for the modular reconfiguration and service‐oriented reuse of decorative elements within manufacturing systems. By integrating semantic understanding with technical innovation, this method not only enhances the precision of pattern extraction but also supports the creative reinterpretation and digital preservation of traditional decorative designs. The study contributes to bridging artificial intelligence with co‐creative practices in design and cultural heritage.