Research Article

Multideep Feature Fusion Algorithm for Clothing Style Recognition

Figure 9

(a) is the precision of the classification of the clothing image style and (b) represents the PR curve. As shown in (a), although the different network models are given different rates, however, the classification accuracy of the sleeves, collars, and patterns of the six types of attributes is higher and the classification accuracy of the three kinds of clothing is low. This is due to the easy distinction between the sleeve length, the collar, and the three kinds of species and the length, the plate, and the clothing are more difficult to distinguish. It can also be seen in (b) that our model has better performance than the other three models.
(a)
(b)