Multilevel and Multiscale Feature Aggregation in Deep Networks for Facial Constitution Classification

Figure 2

The architecture of CNN improved from VGG16.