Creative Product Customization Design Method Based on Image Segmentation and Style Transfer Algorithm

With the growth of personalized consumption demand, customized design of creative products has become an important research direction. However, traditional manual customization methods are inefficient and costly, and it is difficult to meet users’ needs for diversified styles and real-time interaction. To solve this problem, this paper proposes a customized design method that combines deep learning image segmentation and style transfer. The system uses DeepLabV3+ for target area segmentation, and uses StyleGAN2 and CycleGAN for style transfer, enabling users to quickly generate personalized product design solutions. Experimental results show that DeepLabV3+ outperforms U-Net in IoU and Dice scores, and shows higher segmentation accuracy on complex product images. At the same time, StyleGAN2 performs best in style preservation, while CycleGAN achieves a good balance between computational efficiency and style transfer quality. User experience testing further verifies the effectiveness of this method, and the subjects give high scores to its personalized performance and visual quality.