SESR

SESR: Single Image Super Resolution with Recursive Squeeze and Excitation Networks (To appear in ICPR 2018) https://arxiv.org/abs/1801.10319

Quality for scale x4

Trained on div2k, r=4

DataSet/Method	PSNR/SSIM
Set5	32.05/ 0.897
Set14	28.54/ 0.789
BSD100	27.51/ 0.743
Urban100	25.83/ 0.785

Compare with other methods

Requirement

Python 2.7
Pytorch 0.2.0
opencv-python
numpy

Train

python train.py --cuda

Evaluate

python test.py --cuda

Do Super resolution on your own images

python test.py --cuda --mode sr --testdir path_to_your_image

Reference

https://github.com/twtygqyy/pytorch-LapSRN https://github.com/jiny2001/dcscn-super-resolution https://github.com/jmiller656/EDSR-Tensorflow https://github.com/grevutiu-gabriel/python-ssim/blob/master/python-ssim.py