search

cszn / USRNet

Deep Unfolding Network for Image Super-Resolution (CVPR, 2020) (PyTorch)
https://cszn.github.io/
865 stars 115 forks source link

How to get the deblur result? #10

Closed honeytidy closed 4 years ago

honeytidy commented 4 years ago

I run the demo main_test_realapplication.py and can not get the deblur result like you provided example. My result is like this (seems deblur not work): image

cszn commented 4 years ago

Run main_test_table1.py to get the results.

anshulgarg11 commented 4 years ago

I ran the main_test_table1.py but getting similar blur output. I changed the scale factor to 1 before running. Please help.

pablodz commented 4 years ago

Same issue here, maybe the weights?

Left -> LR, Right-> USRNet x4 K11

WhatsApp Image 2020-08-14 at 12 51 08 PM

WhatsApp Image 2020-08-14 at 1 54 35 PM

pablodz commented 4 years ago

Complete Jupyter Notebook to run the model: https://colab.research.google.com/drive/1erpZYSMxN0mAZEcgMmVbnhwoO3IdrTsh?usp=sharing Results attached DOWNLOAD test.zip

I think something it's happening with the model, looks like there're changes. However, they're not doing the work.

cszn commented 4 years ago

Read the paper first. You should set a proper kernel for the LR image.