3D classification example based on DenseNet - Occlusion Sensitivity

[robsver]

Hi all,
I tried downloading and running the 3D classification example based on DenseNet tutorial code on occlusion sensitivity, but unfortunately it didn't generate the same map as the example for me.
This is what I've got so far.
Any ideas?

[ericspod]

Hi @robsver is this 3d_classification/densenet_training_array.ipynb you're referring to? It looks like an interpolation issue perhaps of the image, be sure that you have the most recent versions of MONAI and its dependencies installed and you have run the notebook unmodified.

[robsver]

Hi @ericspod, thanks for your answer. The notebook I was referring to is located in tutorials/3d_classification/densenet_training_array.ipynb, so yes, it is that one. The monai version of my python virtual environment is 1.4.0 (latest) with all its dependencies installed and I ran the notebook unmodified, but it doesn't work for me.

[ericspod]

I'm getting the same results you've got now with current versions of MONAI as well as the dev version. If you look at the tutorial I ran it a long time ago with very old versions of MONAI and its dependencies so I'm afraid a lot has changed since. We should revisit this visualisation feature in the core library, as it seems to me the tutorial is running correctly. Could you please open an issue in the core repo and we'll look into it? Thanks!

[robsver]

Thanks so much for your reply, and sorry for getting back to you a bit late!
I’ve just opened an issue in the core repo in Project-MONAI/MONAI#8428.
Really appreciate your help!
Thanks again!