Code implementation and questions about the article Action Machine: Toward Person-Centric Action Recognition in Videos

[aoteman233]

Excuse me, I saw your article on human motion recognition on the Internet and was very interested. I have always wondered whether I3D as a 3D convolutional network can be integrated with 2d convolutional networks such as vgg. I would like to know the specific connection operation between I3D and 2D CNN in your article. Would you mind sharing this part of the code?

thank you very much!