Fig. 5From: Improved two-stream model for human action recognitionThe structure of VGG16Back to article page