F-Retinex-Net(paper)

Feature Based Deep Retinex for Low-Light Image Enhancement

This is a Tensorflow implement, fork from Retinex-Net

Network Structure:

Requirements

1. Python
2. Tensorflow >= 1.5.0
3. numpy, PIL

Testing Usage

you can just see some demo cases by

python main.py --phase=test --test_dir=./data/test/low

, the results will be saved under ./test_results/.

Some enhanced results:

Training Usage

First, download training data set from Dataset. Save training pairs of our LOL dataset under ./data/our485/, and synthetic pairs under ./data/syn/.

Moving some data from our485 to eval15:

cd data; python mv2eval.py

, all eval files: ./data/eval15/evel50.txt

Then, just run

sh train.sh

model_end2end.py is end to end method, use it by uncommenting main.py.