I’m trying to use SlowFast50 for video classification, however, I don’t get how properly pass the input to the pre-trained network. In fact, I’ve already extracted the desired number of frames and put it into
segment = torch.stack(frames,0).to(device) where
frames is a list and each element is a tensor already normalized between 0 and 1.
Can I pass then
segment directly into
model(input) without performing the step
Sorry but I am new into the Slowfast networks.