search

xinntao / EDVR

Winning Solution in NTIRE19 Challenges on Video Restoration and Enhancement (CVPR19 Workshops) - Video Restoration with Enhanced Deformable Convolutional Networks. EDVR has been merged into BasicSR and this repo is a mirror of BasicSR.
https://github.com/xinntao/BasicSR
1.49k stars 318 forks source link

想咨询论文中关于PCD的溶解实验 #168

Closed ryanxingql closed 4 years ago

ryanxingql commented 4 years ago

您好！EDVR文中对PCD的溶解实验有这么一幅图：

image

而PCD是对feature进行的alignment，而不是对frame进行alignment。请问28和30帧的灰度图又是如何得到的呢？

感谢您的解答，祝您工作顺利！

xinntao commented 4 years ago
  1. 灰度图。当我们得到warped features后，灰度图只是visualization。选择一个通道， 将其归一化到[0， 1], 然后当作图片来visualization即可。
  2. Feature的光流。同理，我们选取feature的一个通道, 将其归一化到[0， 1], 当作图像输入到flow net, 可以得到光流。
wenchen4321 commented 3 years ago

Hello, How did you get Flow Magnitude from the optical flow map?