deepfrAI

An ESRGAN model to deep fry images.

How to use this?

Download the model from the releases page, chaiNNer is the easiest way to use the trained model. Other tools that work with ESRGAN models should be compatible as well.

Used ressources

• Dataset https://www.kaggle.com/datasets/sayangoswami/reddit-memes-dataset
• traiNNer: https://github.com/victorca25/traiNNer
• Training guide: https://rentry.org/How2ESRGAN

Training

Preparing the dataset

1. Download and extract the dataset
2. Deep fry the data set python deepfry.py, edit deepfry.py if necessary
3. Manually move the images into the correct directories under ./dataset
   • Most original images to ./dataset/train/lr
   • The corresponding deep fried images to ./dataset/train/hr
   • Some original images to ./dataset/val/lr
   • The corresponding deep fried images to ./dataset/val/hr

Training

cd traiNNer/codes
python3 train.py -opt train_sr.yml

I trained the model for 40000 iterations.