Firstly, this project was implemented using Python 2.7, so at the begining you should check your Python version.

The Python package needed:

• TensorFlow(GPU version)
• Numpy
• SciPy
• matplotlab
• PIL
• pickle
• argparse
• glob

Secondly, we use MATLAB(Linux version) to preprocess the training data(because when we generate the training data from image, using MATLAB is much faster than using openCV on Python). So make sure you have already installed MATLAB on your server.

Finally, you can get the training data from: data

• Download the training and test data from here
• Unzip the data.zip, unzip 291.zip and Set14.zip inside.
• Copy the 291 and Set14 directory to "/data" directory.
• Run generate_train.m and generate_test.m to generate training and test data.

Run "train.py" to train. If you want to start from a checkpoint, using "py -2 train.py --model_path [your checkpoint path]"

Run "python test.py"

Run "python plot.py"

Run python increase_resolution.py --img [your-image-path]

If you want to use the network without training, download this model, the result may be not very good since it just a test model.