UnblurNet

Removes different types and intensities of blurs from images

Setup a virtual environment

• Install python3-virtualenv pakage system-wide sudo apt-get install python3-virtualenv
• Create a virtual environment inside the project directory python3 -m venv .venv
• Activate your venv source .venv/bin/activate

Install project dependencies

pip install -r requirements.txt

Get the dataset from kaggle

After downloading the dataset setup the project directory in such a way

UnblurNet
├───inputs
│   ├───defocused_blurred
│   ├───gaussian_blurred
│   ├───motion_blurred
│   └───sharp
├───outputs
│   └───saved_images
└───src

Testing the model

• Currently the train.py script doesn't use the blurred images provided in the dataset.
• You can create a new directory with gaussian blurring using the utils/add_gaussian_blur.py script python3 utils/add_gaussian_blur.py
• Now we can run the src/train.py script python3 src/train.py --epochs 100