Skip to content
A Pytorch implement of the paper 2018 NTIRE No.1 paper 《Wide Activation for Efficient and Accurate Image Super-Resolution》
Python
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
checkpoint
samples
README.md trying to write a good readme Jan 24, 2019
args.py Add files via upload Jan 3, 2019
main.py
psnr.py Add files via upload Jan 3, 2019
train.py Add files via upload Jan 3, 2019
utility.py Add files via upload Jan 3, 2019
wdsr_b.py
记录psnr.numbers Add files via upload Jan 10, 2019

README.md

2018.12视频通信大作业

A Pytorch implement of NTIRE2018 No.1 network WDSR https://arxiv.org/abs/1808.08718v1
Dataset: DIV2K 2017 https://data.vision.ee.ethz.ch/cvl/DIV2K/

DATA 
├── HR  
└── LR

Training data is augmented with random horizontal filp and rotations, check utility.py and rewrite class SRdataset!

How to train

Delete & make new

vim ./loss.log
mkdir ./samples
mkdir ./checkpoint

GPUs are needed for training

python main.py --cuda

How to test

Test method

700x700 HR image and its LR counterpart are randomly cropped from every image in DIV2K Validset
Calculate the mean PSNR of HR image and Image Restored by network

make correspond empty folder to store samples before test

mkdir ./foldername/

change samples save_path and model to restore in psnr.py

python psnr.py

Specific description of given samples, checkpoint as well as test results can be found in .numbers file ^_^

Result

Truth

LR

Output

You can’t perform that action at this time.