Second-hand clothing classification algorithm based on featurefusion and attention

This study suggests a second-hand clothing categorization method based on enhanced residual networks to enhance the effect of second-hand clothing retrieval and encourage clothing transactions on second-hand platforms. This study first gathers image data on used garments. Web crawlers are utilized to gather internet photos of second-hand clothes to train the network model, while camera equipment is used to take pictures of second-hand clothing. The resulting images are then used to assess the network model’s categorization accuracy. The next step is to construct a classification model based on ResNet50, add an attention mechanism, and carry out feature extraction in stages. Finally, the developed classification model’s performance is assessed and contrasted with other approaches. The experimental findings demonstrate that this strategy outperforms previous methods in terms of classification accuracy on the self-built dataset and DeepFashion dataset, reaching 79.69% and 82.22%, respectively. Additionally, the sorting and recycling of used clothing is greatly assisted by this method.