pytorch-i3d
pytorch-i3d copied to clipboard
How to get per frame embeddings given 1 video
Seems for i3d I need to fill in 16 frames to get 1 embedding because of the strides and ave pool. Given a video of length T, how to construct the 16 frames to get T embeddings?
Seems for i3d I need to fill in 16 frames to get 1 embedding because of the strides and ave pool. Given a video of length T, how to construct the 16 frames to get T embeddings?
Hello, I have encountered the same problem. Have you solved this problem?
同问,请问有什么好的方法