Deep Ensemble Learning for Human Action Recognition in Still Images

Figure 4

The specific structure of VGG16_NCNN. It incorporates three modules: the VGG16_base module, NCNN module followed by the GAP layer, and the classifier module.