GitHub - huyangdi/Deep-Semantic-Face-Deblurring: Deep Semantic Face Deblurring

Branches Tags

Name		Name	Last commit message	Last commit date
Latest commit History 6 Commits
blur_image		blur_image
deblur_image		deblur_image
matconvnet-1.0-beta22		matconvnet-1.0-beta22
real_blur_image		real_blur_image
.gitattributes		.gitattributes
DL_deblur_net18.m		DL_deblur_net18.m
DL_deblur_net18.m~		DL_deblur_net18.m~
Sum.m		Sum.m
blur.png		blur.png
example.png		example.png
example_deblur.png		example_deblur.png
kernel_27_(1).mat		kernel_27_(1).mat
main_deblur18.m		main_deblur18.m
main_deblur_all.m		main_deblur_all.m
main_deblur_all.m~		main_deblur_all.m~
move.m		move.m
net_G_P_S_F.mat		net_G_P_S_F.mat
net_G_P_S_F_GAN.mat		net_G_P_S_F_GAN.mat
net_P_P_S_F.mat		net_P_P_S_F.mat
net_P_P_S_F_GAN.mat		net_P_P_S_F_GAN.mat
readme.txt		readme.txt
savemodel.m		savemodel.m

Repository files navigation

The code is implementation code for the following paper: 

Ziyi Shen, Wei-sheng Lai, Tingfa Xu, Jan Kautz and Ming-Hsuan Yang 
Deep Semantic Face Deblurring
IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2018

Requirements 
MATLAB (We test with MATLAB R2016a on Windows 10)


Test Pre-trained Models

Compile matconvnet:
>> cd matconvnet-1.0-beta22/matlab
>> vl_compilenn('enableGpu', 1, 'enableCudnn', 1)
>> cd ../../

We provide the matconvnet in,
./DL_deblur_net./matconvnet-1.0-beta22./matlab


Run  'main_deblur18.m' to test the example.png

or 

You also can run 'main_deblur_all.m' to test the blur images in ./blur_image and ./real_blur_image.

In addition, the whole testing datasets have been released on our project website
https://sites.google.com/site/ziyishenmi/cvpr18_face_deblur


We provide two models here.

1. net_G_P_S_F_GAN and net_P_P_S_F_GAN
Our model with L1 loss + parsing Loss + structure loss +feathure loss

2. net_G_P_S_F and net_G_P_S_F

Our model with L1 loss + parsing Loss + structure loss + feathure loss + adversarial loss