Learning image aesthetics by learning inpainting

Overview of the project

Architecture

Methods of image inpainting

Ways to apply the masking are in inpainting/datasets.py

Getting started for Self-supervised learning with image inpainting

Step 1) Installation of packages with pip

pip install -r requirements

Step 2) Change of paths

May change the path to dataset, pre-trained models as well as the architecture of models to the one you want to.

Step 3) Start training

python image_inpainting.py

Aesthetics quality assessment task training on AVA

Step 1) Download AVA

link

Step 2) Download CSV file (AVA1 and AVA2)

gdrive link

Step 3) Change of paths

May change the path to dataset, pre-trained models as well as the architecture of models to the one you want to.

Step 4) Start training

python main.py

Acknowledgement

Parts of this code borrows from kentsyx and eriklindernoren repositories.