Nav apraksta

SUN Hao f452a354b5 migrate to tenforflowf 2.0 4 gadi atpakaļ
data fce786fdcf release code 6 gadi atpakaļ
figs 5b5b4778db Add files via upload 6 gadi atpakaļ
model fce786fdcf release code 6 gadi atpakaļ
LICENSE fdc15ebc17 Create LICENSE 5 gadi atpakaļ
README.md 2d94be17c0 Update README.md 5 gadi atpakaļ
main.py f452a354b5 migrate to tenforflowf 2.0 4 gadi atpakaļ
model.py f452a354b5 migrate to tenforflowf 2.0 4 gadi atpakaļ
utils.py fce786fdcf release code 6 gadi atpakaļ

README.md

RetinexNet

This is a Tensorflow implement of RetinexNet

Deep Retinex Decomposition for Low-Light Enhancement. In BMVC'18 (Oral Presentation)
Chen Wei*, Wenjing Wang*, Wenhan Yang, Jiaying Liu. (* indicates equal contributions)

Paper, Project Page & Dataset

Requirements

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

Testing Usage

To quickly test your own images with our model, you can just run through

python main.py 
    --use_gpu=1 \                           # use gpu or not
    --gpu_idx=0 \
    --gpu_mem=0.5 \                         # gpu memory usage
    --phase=test \
    --test_dir=/path/to/your/test/dir/ \
    --save_dir=/path/to/save/results/ \
    --decom=0                               # save only enhanced results or together with decomposition results

Or you can just see some demo cases by

python main.py --phase=test

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

Training Usage

First, download training data set from our project page. Save training pairs of our LOL dataset under ./data/our485/, and synthetic pairs under ./data/syn/. Then, just run

python main.py
    --use_gpu=1 \                           # use gpu or not
    --gpu_idx=0 \
    --gpu_mem=0.5 \                         # gpu memory usage
    --phase=train \
    --epoch=100 \                           # number of training epoches
    --batch_size=16 \
    --patch_size=48 \                       # size of training patches
    --start_lr=0.001 \                      # initial learning rate for adm
    --eval_every_epoch=20 \                 # evaluate and save checkpoints for every # epoches
    --checkpoint_dir=./checkpoint           # if it is not existed, automatically make dirs
    --sample_dir=./sample                   # dir for saving evaluation results during training

Tips:

  1. The model is quite small, so it will take just minutes to finish the training procedure if you are using GPU. For people who are using CPU, it is also affordable.
  2. The enhancement performance is highly dependent on training parameters. So if you change the default parameters, you might get some weird results.

### Citation ###

 @inproceedings{Chen2018Retinex,
  title={Deep Retinex Decomposition for Low-Light Enhancement},
  author={Chen Wei, Wenjing Wang, Wenhan Yang, Jiaying Liu},
  booktitle={British Machine Vision Conference},
  year={2018},
  organization={British Machine Vision Association}
}