Skip to content
Dynamic Scene Deblurring Using Spatially Variant Recurrent Neural Networks[matcaffe]
Branch: master
Clone or download
Latest commit 8522113 Mar 8, 2019
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
cmake Add existing file Aug 19, 2018
examples Add existing file Aug 19, 2018
include Add existing file Aug 19, 2018
matlab Delete 000047_125.png Mar 8, 2019
scripts Add existing file Aug 19, 2018
src Add existing file Aug 19, 2018
tools Add existing file Aug 19, 2018
CMakeLists.txt Add existing file Aug 19, 2018
LICENSE Initial commit Aug 19, 2018
Makefile Add existing file Aug 19, 2018
Makefile.config
caffe.cloc Add existing file Aug 19, 2018
contributors.txt Add existing file Aug 19, 2018
readme.txt Update readme.txt Mar 8, 2019

readme.txt

This package contains the code of the image deblurring algorithm described in the paper: 
Jiawei Zhang, Jinshan Pan, Jimmy Ren, Yibing Song, Linchao Bao, Rynson W.H. Lau, Ming-Hsuan Yang, "Dynamic Scene Deblurring Using Spatially Variant Recurrent Neural Networks", CVPR 2018. 
----------------
How to use
----------------
1. Please install matcaffe by
(1) compile caffe: make all -j8.
(2) compile matcaffe: make matcaffe.
2. Put training data into: ./matlab/train/data. There are a few images in that folder already.
3. Run train.m in ./matlab/train for training
4. Run test.m in ./matlab/test for testing
5. Download caffe model for testing in https://drive.google.com/file/d/1uZ9QcUR1R7k5nFeCj4D7IZopJYr5b37Q/view?usp=sharing

Thanks Sifei Liu to provide her matcaffe code from [1] and spatial rnn implementation from [2]
[1] Sifei Liu and et. al. Multi-Objective Convolutional Learning for Face Labeling, CVPR 2015
[2] Sifei Liu and et. al. Learning Recursive Filters for Low-Level Vision via a Hybrid Neural Network, ECCV 2016 
You can’t perform that action at this time.