TensorFlow implementation of "WESPE: Weakly Supervised Photo Enhancer for Digital Cameras"
- Download the pre-trained VGG-19 model and put it into
vgg_pretrained/
folder - Download DPED dataset (patches for CNN training) and extract it into
dped/
folder.
- python train_model.py model= or sh train.sh
test PSNR for iphone is about 18 dB, similar to the original paper, 18.11 dB