MECLabTUDA / M3d-Cam

MIT License
310 stars 40 forks source link

3D segmentation Demo Activation Maps Visualisation #19

Open KristoferLintonReid opened 2 years ago

KristoferLintonReid commented 2 years ago

Hi,

I have worked through the 3D segmentation demo on using nnUNet. Can you please add some code to visualize the activation maps like what you have added in the table column 3?

I may be misunderstanding is what looks like segmentation masks in the .nii.gz files in the inference_results folder the attention maps? Otherwise, can you add some example code to visualize maps from the .pkl file?

Thanks

Karol-G commented 2 years ago

Hi,

I have worked through the 3D segmentation demo on using nnUNet. Can you please add some code to visualize the activation maps like what you have added in the table column 3?

I am not aware that there are python libraries to visualize 3D data / 3D attenion maps in python. You need to download the attention_map.nii.gz files and open them with an nii.gz viewer like ITK-Snap or MITK.

I may be misunderstanding is what looks like segmentation masks in the .nii.gz files in the inference_results folder the attention maps? Otherwise, can you add some example code to visualize maps from the .pkl file?

In the nnU-Net demo the attention maps are saved in the folder attention_maps. The folder inference_results is probably a folder used by the nnU-Net.

Best Karol