search

sniklaus / pytorch-pwc

a reimplementation of PWC-Net in PyTorch that matches the official Caffe version
GNU General Public License v3.0
621 stars 123 forks source link

Use correlation as loss #15

Closed bkvie closed 5 years ago

bkvie commented 5 years ago

correlation(image1, image2) outputs a multidimensional tensor, would it be possible to produce a scalar that can be used as a loss. I.e deform image 1 and minimise negative correlation between image1, image2.

summing up over all dimensions doesn't work.

sniklaus commented 5 years ago

Very interesting idea! I am not an expert in optical flow estimation, so my insights into this behavior are limited. I would thus encourage you to bring this up in the official repo: https://github.com/NVlabs/PWC-Net