clarity of dataset directory structure, dependencies, model_weights

[zz397]

Hi,

I am very interested in your work "SPAN: Spatial Pyramid Attention Network for Image Manipulation Localization". I want to run your evaluation script to better understand the paper. Can you provide a bit more details about the following:

Let's use Columbia dataset as an example.

1. what should be the Columbia_image_new.list file? It's missing in the repo
2. what are the dependencies to run the scripts? what version of tensorflow, keras ... should be installed?
3. which is the model corresponding to the provided weights(PixelAttention32.h5)? is it the weights of the pretrain-only model from the paper?

Thank you!