Multideep Feature Fusion Algorithm for Clothing Style Recognition

Figure 1

Multifeature extraction and fusion process. Firstly, using the improved detected method to extract the global area, main part, and part areas of the input image. Secondly, according to the results of the method, the multicategory areas is inputted to the improved residual network. And the residual network separately outputs three 128 dimensional features of global area, main part, and parts areas. Finally, the three category features are effectively fused by the multicategory feature fusion technology.