Pattern Redesign Imitating Ethnic Clothing Color Styles via Palette-guided GAN

The digital regeneration of ethnic clothing colors is pivotal in fashion design. With advancements in information technology, machine learning techniques are increasingly being applied to the design of ethnic clothing. However, existing color transfer techniques often focus on single-color blocks, overlooking the complexity of color layering and proving inadequate for large-scale image colorization. To address these challenges, we propose an automatic colorization method guided by a color palette. The generative network is enhanced by incorporating a primary coloring network and a conditional network based on generative adversarial network principles. The conditional network integrates color palette inputs, while the primary network processes image data, enabling palette-guided colorization of non-colored patterns. Huber loss is introduced to the generator’s loss function to improve colorization precision. Experimental results demonstrate that our algorithm surpasses existing methods in structural similarity and peak signal-to-noise ratio values, with an average colorization time of less than 0.2 seconds per pattern. The proposed solution effectively supports the digital, intelligent mass design of ethnic clothing coloration, ensuring interactivity and style consistency, and offers valuable insights for fashion product colorization.