codersadis / content-aware-cnn-loop-filter-for-hevc Goto Github PK

Hey
the complete code is not there. Can you please update it?

Thank You

How did you generate the HEVC compressed data? e.g. qp22.

How do you extract the images in loop? Can you share the code?

Thx

Hi, I read your paper "Content-Aware Convolutional Neural Network for In-Loop Filtering in High Efficiency Video Coding" recently and got some questions.

The first one is, you said the dropout probability is 1. Does it mean the keep probability is 1, or the drop probability is 1. Both look strange.

Another one is, for the first "connection unit" in the network, the input x and output y have different feature map numbers (x is [?, 38, 38, 1] and y is [?, 38, 38, 64], is that right?). How dose the dropout connection change the feature map number of x, so that the output of dropout connection is consistent with the output of convolution layer dimension?

Thanks and looking forward to your reply.