Dynamic style transfer for interior design: An IoT-driven approach with DMV-CycleNet

Abstract

With the rise of IoT in smart home environments, interior design increasingly relies on style transfer to create personalized and adaptive spaces that respond to user preferences and environmental conditions. However, existing methods often struggle to preserve spatial structure, detail, and adapt dynamically to changes in complex indoor scenes. To address these limitations, we propose DMV-CycleNet, an integrated model combining high-precision segmentation, depth estimation, and IoT-driven adaptive style transfer, leveraging real-time environmental data to refine visual outputs. By incorporating IoT feedback, DMV-CycleNet dynamically adjusts style elements to suit varying lighting, spatial configurations, and user-defined preferences, enhancing both adaptability and personalization in interior design. Experimental results demonstrate that DMV-CycleNet outperforms baseline models in SSIM, PSNR, and FID metrics across diverse artistic styles, effectively preserving structural coherence and object boundaries. This model advances the practical application of style transfer in IoT-enabled smart home environments, contributing to the development of truly responsive and intelligent interior spaces.

Keywords

• Interior design
• Style transfer
• IoT
• Depth estimation
• Segmentation
• Adaptive style transfer