Does gradCAM work on 3D image inputs?

I have 3D MRI volume in a 256x256x49 uint8 array. The same is the input expectation of my model.

[class,score]=classify(trainedNetwork_1,MRI)

The above statement runs fine, with output:

class =

categorical

MS

score =

1×2 single row vector

0.6391 0.3609

but the following function throws an error:

gradCamMap=gradCAM(trainedNetwork_1,MRI,class);

Error using gradCAM (line 144)

Index exceeds the number of array elements. Index must not exceed 2.

What am I doing wrong here?

Answer 1

[Hitesh]

Hi @Khuhed Memon

From what I understand is that if your model has multiple classes, ensure that the index you are passing to "gradCAM" corresponds to the correct class index. You need to convert the categorical class label to an index. Kindly refer to the below code for an example:

classIndex = find(class == categories(class)); % Example conversion to index
% Call the gradCAM function
gradCamMap = gradCAM(trainedNetwork_1, MRI, classIndex);

For more information on "gridCAM", refer to this MATLAB documentation:

https://www.mathworks.com/help/deeplearning/ref/gradcam.html

If the issue persists, could you please share the code file where you are encountering this error? So that I can investigate the issue.