A TensorFlow implementation of CVPR 2018 paper Residual Dense Network for Image Super-Resolution.
Official implementation: Torch code for our CVPR 2018 paper "Residual Dense Network for Image Super-Resolution" (Spotlight)

• TensorFlow-1.10.0
• Numpy-1.14.5
• OpenCV-2.4.9.1
• PIL-3.1.2
• h5py-2.6.0

Download DIV2K training data from here.
Extract and place all the images in RDN-TensorFlow/Train/DIV2K_train_HR.

python main.py

python main.py --is_train=False

If you want to use the resize function in MATLAB when generating training data and testing images as the pretrained model used, you need to install MATLAB API for Python, and run the script with option --matlab_bicubic=True.

If you want to take an original image as the input of RDN directly, you could run the script like python main.py --is_train=False --is_eval=False --test_img=Test/Set5/butterfly_GT.bmp.

• kweisamx/TensorFlow-ESPCN