I'm interested in understanding how the network learns the intermediate steps (depth, optical flow...).
In #31 you recommend checking the PWC-Net repository. However, from the paper I understand that you keep training that module (model) and indeed set different learning rates for each.
Could you please guide me on how did you visualize the result of the depth and optical flow. Or maybe what's the specific layer/children that output this information?
Hi!
Congrats for this work, it's very cool.
I'm interested in understanding how the network learns the intermediate steps (depth, optical flow...).
In #31 you recommend checking the PWC-Net repository. However, from the paper I understand that you keep training that module (model) and indeed set different learning rates for each.
Could you please guide me on how did you visualize the result of the depth and optical flow. Or maybe what's the specific layer/children that output this information?
Thank you very much.