The output of pretrained model is a list with 32 tensors #43
Description
When I used pretrained model in visualize_flow.py, everything is alright, compute_flow() function outputs a [1,2,H,W] shaped tensor.
But when I use the model in my own project, something strange happened. The compute_flow() function outputs a list which contains 32 tensors, which I guess are flows. But I only input two shape-correct frames. Have you ever met with this situation?
