Skip to content
inzva AI Projects #2 - Image Restoration Project
Branch: master
Clone or download
Latest commit 849b15a Jul 8, 2019
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
dataset Initial commit May 31, 2019
result_Sony Initial commit May 31, 2019
LICENSE Initial commit Mar 19, 2019
README.md Update README.md Jul 8, 2019
test_Sony.py Initial commit May 31, 2019
train_Sony.py Initial commit. May 31, 2019

README.md

Image-Restoration Team

Contributors (All equally contributed):

  • Ahmet Melek
  • Onur Boyar
  • Furkan Gürsoy
  • Burak Satar

We restore very dark images to high quality and visible images.

Here is an example from the reference paper: Here is an example from the reference paper

Our purposes on this project are:

1- Reproduce the results of Learning to See in the Dark project, as can be seen here: https://github.com/cchen156/Learning-to-See-in-the-Dark

2- Obtain results faster via optimization of the code.

3- Trying to have better results with modifications. (optional goal)

4- Testing different architectures on this problem. (optional goal)

inzva AI Projects #2 - Image Restoration Project

Instructions

1- Paths and hyperparameters can be set at the top of test_Sony.py and train_Sony.py files.

2- The files will be read from respective input and ground truth directories.

3- The size of the deep neural network will be decided based on hyperparameters.

4- Training and test sets are generated based on the first characters of the filenames. Please refer to the code for specific implementation.

5- Output images and trained models will be saved in result and checkpoint directories.

6- For both training and test; epoch, loss, time information are printed during execution.

Let us know if you spot any error or have any suggestions.

You can’t perform that action at this time.