Image Denoising Using UNets and Perceptual Loss

The project aims at proposing a model that does automatic image denoising using a UNet architecture and Perceptual loss.

Model Architecture

The core of the denoising model is the U-Net architecture, which consists of an encoder-decoder structure. The encoder compresses the input image into a lower-dimensional feature space, capturing essential features, while the decoder reconstructs the image from this compressed representation.

Directory Structure

Note : The model was trained on kaggle P100 GPU. There in some places in model.ipynb file you might find kaggle/working as directory.

Image-Denoising/
├── denoising-skipconnections (1).ipynb (model with skip connections, couldn't train due to lack of computation)
├── model.ipynb (without skip connections, trained)
├── requirements.txt
├── README.md
├── test/
│   ├── high/
│   └── low/
├── models/
│   ├── 10_unet_ploss_vgg19.pth
│   ├── 15_unet_ploss_vgg19.pth
│   ├── 20_unet_ploss_vgg19.pth
│   ├── 25_unet_ploss_vgg19.pth
├── 25_unet_ploss_vgg19.pth
├── main.py
└── running_on_test_data/
     ├── 15_unet_ploss_vgg19.pth
     ├── experiments.ipynb
     ├── testData_prep.ipynb
     ├── test_psnr.ipynb
     └── images/
          ├── ground_truth/
          ├── high/
          └── low/

Installation

To deploy this project run

git clone https://github.com/Ayush-Singh677/Image-Denoising.git
cd Image-Denoising

Create a virtual environment and activate it:

python -m venv venv
source venv/bin/activate  # On Windows, use `venv\Scripts\activate`

Install the required dependencies:

pip install -r requirements.txt

Usage

Insert images images that you want to denoise in /test/low Run the following command

python3 main.py

Your denoised images will be saved to /test/low