Pytorch Implementation of Noise2Noise & Module Implementation

Preparation

Download and extract data folder from link, and place it in the root folder of this repo.

Tests Pass Screenshots

Using GPU

Using CPU

Special Thanks

We thank VITA Lab for providing clusters on izar for training our models in Miniproject 2!

Follow up fixes

1. Fixed Sequential backpropagation and parameter saving to resolve potential issues that could arise when building such a structure.

self.model = Sequential(
    Sequential(
        Conv2d(3, 32, 3, stride=2, padding=2),
        ReLU(),
        Conv2d(32, 64, 3, stride=2, padding=2),
        ReLU()),
    Upsampling(2, 64, 32, kernel_size=4, stride=1),
    Sequential(
        ReLU(),
        Upsampling(2, 32, 3, stride=1, kernel_size=3),
        Sigmoid())
).to(self.device)