/rise-testing

The runs of the notebook provided with the RISE paper

Primary LanguageJupyter NotebookMIT LicenseMIT

What is this repo about?

It checks what exactly the RISE notebooks do.

How to launch the notebooks?

  1. Create the conda environment with the given .yml file.
  2. The environment will download the GPU versions of tensorflow and pytorch, which is around 4GB total, so it will take some time.
  3. If you want to run the Saliency notebook, you need to download the Image Net 2012 validation dataset. And put the pictures into whatever_folder/whatever_class/actual_pictures so the image loader from the notebook works as intended.
  4. Now you can run jupyter notebook and open the notebooks in it.

What is different?

Compared to the original repository,
In Easy_start, keras became tensorflow.keras because it has become a part of the tensorflow package some time ago.
In Saliency, the image folder and gpu batch size were changed.