This is a fast adaptation to tf2 of deep photo enhancer by @nothingio https://github.com/nothinglo/Deep-Photo-Enhancer.

The data used to train the model is the MIT adobe dataset.

Locations for the training files and test files can be set in the data.py file. For moredetails you can look at the original repository.

To run inference with the saved pretrained model run python predict.py

This will work only for image of maximum size 512.