How to plot a ROC curve for CNN?

I have trained a CNN in Matlab 2019b and I have a trainednet.mat file which has the trained CNN. I am able to classify images using the classify(net,im) function but I am unsure of how to generate a ROC curve. I have seen the perfcurve() function but I am unsure of how to get all the inputs the function needs. Any help would be appreciated.

Answer 1

[Jalaj Gambhir]

Hi,

Have a look at the solution given in this post:

https://www.mathworks.com/matlabcentral/answers/373849-roc-curve-alexnet-cnn

perfcurve requires 3 input arguements, the ground truth labels of each image, the predicted scores by classify function, and the class labels that is considered to be positive. Look at the example in the perfcurve documentation link shared above,

Hope this helps!