/PyTorch-Noise2Noise

Project for EE-559 Deep Learning @ EPFL: Pytorch Implementation of Noise2Noise & Building Modules of Neural Network from Scratch

Primary LanguagePython

EE-559 Project: Learning Image Restoration without Clean Data

The README outlines the requirements, project structure and the way to reproduce our final results. For more details, please check our report.

Introduction

This project aims to implement a Noise2Noise model. A Noise2Noise model is an image denoising network trained without a clean reference image. The original paper can be found here.

The project has two parts, focusing on two dierent facets of deep learning. The rst one is to build a network that denoises using the PyTorch framework, in particular the torch.nn modules and autograd. The second one is to understand and build a framework, its constituent modules, that are the standard building blocks of deep networks without PyTorch's autograd.

Requirements

  • Python 3.7
  • numpy 1.21.4
  • PyTorch 1.10.0+cu102
  • collection
  • pathlib

Project Structure

Here is an overview of the architecture of our project:

|--Miniproject_1
	|--__init__.py
	|--model.py
	|--bestmodel.pth
	|--Report_1.pdf
	|--others
		|--__init__.py
		|--config.py
		|--datasets.py
		|--networks.py
		|--train_example.py
		|--utils.py
|--Miniproject_2
	|--__init__.py
	|--model.py
	|--bestmodel.pth
	|--Report_2.pdf
	|--others
		|--helpers.py
		|--NNUsampling_conv.py
|--data
	|--train_data.pkl
	|--val_data.pkl
|--results
|--README.md

How to run

We used the GPU, provided by VITA lab, to run our model. It is also possible to run our model with CPU, but it would be time-consuming.

To reproduce our final results or train from scratch (You can use the data provided in this repo and skip step 1&2, or use your own data):

  1. Upload your own data and make sure they are in directory ./data/.

  2. Split the data into training set and validation set, and name them as in project structure.

  3. Play with the model using ./Miniproject/others/train_example.py:

    1. use default model parameters to train and evaluate the model from scratch by running:
    cd Miniproject/others
python train_example.py
    1. If you have pretrained checkpoint, set resume argument as True and place the checkpoint in directory ./Miniproject/bestmodel.pth.

  4. In ./Miniproject_2/model.py, we provide self-implemented modules widely used in Convolutional Neural Network (CNN), such as Conv2d, Upsampling, SGD, and MaxPool2d layers. The forward and backward processes are self-contained.

Results