Keras implementation of Efficient Sub-Pixel Convolutional Neural Networks for super-resolution. The original paper is Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network.
- Tensorflow 1.12.0
- keras 2.2.4
- python 2.7.16
- numpy
- opencv
- skimage
This code requires keras with a tensorflow backend.
run python download_datasets.py to download the datasets BSD300 to the data folder, BSD300 contains training and test datasets
For training, run
python train.py --scale_factor=2
For testing, run
python test.py --scale_factor=2
the Super-Resolution images are in the result folder
When training the models with --scale_factor=1, it can be used for image enhancement such as deblur or mosaic Elimination and so on based on your train datasets, and you need to modify the data.py to meet your functional needs. For example, When scale_factor=1 this project change to a deblur net
Train for image enhancement, run
python train.py --scale_factor=1
Test for image enhancement, run
python test.py --scale_factor=1