SRN-DeblurNet

For training,I strictly follow all configuration of the original paper.

To test PSNR, I split test set of GOPRO datset into two half, one for validation and the other for test, and get 29.58db PSNR ( 30.26db reported in the original paper).

A pretrained model is provided. It is trained on GOPRO's blurry images without gamma correction. GOPRO dataset

I tryied to ultize this model to augment real world face images, but found it doesn't generalize well.

Any discussion or correction is welcomed.

Reference

Scale-recurrent Network for Deep Image Deblurring

Deep Multi-scale CNN for Dynamic Scene Deblurring

Name		Name	Last commit message	Last commit date
Latest commit History 7 Commits
.gitignore		.gitignore
LICENSE		LICENSE
README.md		README.md
SRNDeblurNet_epoch1999.pth		SRNDeblurNet_epoch1999.pth
conv_lstm.py		conv_lstm.py
data.py		data.py
layers.py		layers.py
log.py		log.py
network.py		network.py
test.py		test.py
test_save.py		test_save.py
train.py		train.py
train_config.py		train_config.py
try.py		try.py
utils.py		utils.py

Folders and files

Latest commit

History

LICENSE (adsbygoogle = window.adsbygoogle || []).push({});

network.py (adsbygoogle = window.adsbygoogle || []).push({});

Repository files navigation

SRN-DeblurNet

Reference

LICENSE

network.py