A Furniture Image Style Classification Model Based on the Fusion of ResNet and Swin Transformer

Furniture style describes the distinctive visual features of furniture, playing a significant role in real-world interior decoration. Product classification based on aesthetic attributes such as style is an important complement to traditional classification techniques. This paper explores the features of furniture styles and investigates the problem of furniture style classification. Unlike traditional object classification, furniture style classification aims to categorize different furniture items based on their “style” that describes their appearance, rather than the more functionally relevant “type” (e.g., bed, table, etc.). To address the issue of poor classification performance in furniture image style models, we combine ResNet and Swin Transformer networks, integrating local and global features of furniture images through Attention Feature Fusion (AFF) module. This approach significantly enhances the accuracy of furniture style classification.