Computer System and Networks Group Project
Code and some images takes from: https://github.com/tensorlayer/srgan
This repo does NOT include all the necessary data and files to train the network. Images and model files would take up way too much space and training would take ages anyways.
In case you want to go through all the effort you will also need the vgg19 pretrained model:
Download the VGG19 model and place it in the folder models/
The notebook.ipynb shows the evaluation of an input image
When running the evaluation do NOT just execute all code cells. Some are just for explanation and will fail! execute just the imports and the ones below section 2 Training / Evaluation.
Depending on the installed versions of the dependencies, executing the evaluation or the training locally can be difficult.
If errors occur in the files mentioned below, then make the following modifications:
Modify multiple lines of python-installation/site-packages/tensorlayer/files/util.py
layer_names = [n.decode('utf8') for n in f.attrs["layer_names"]] or similar
to layer_names = f.attrs["layer_names"]
Modify multiple lines of python-installation/site-packages/tensorlayer/models/core.py
if isinstance(check_argu, tf_ops._TensorLike) or tf_ops.is_dense_tensor_like(check_argu):
to if isinstance(check_argu, (tf.Tensor, tf.SparseTensor, tf.Variable)) or tf_ops.is_dense_tensor_like(check_argu):