ROC curve AlexNet CNN

How to calculate the ROC curve using AlexNet CNN from Matlab? I have two class.

Answer 1

[Gledson Melotti]

cgt = double(testeImagesLabels); clabel = double(Test_predict); cscores = double(Probability);

figure(2) [X,Y,T,AUC,OPTROCPT,SUBY,SUBYNAMES] = perfcurve(cgt,cscores(:,1),1); plot(X,Y,'k');

[Win Sheng Liew]

May I know what is your testeImagesLabels,Test_predict and Probability?

[Gledson Melotti]

testeImagesLabels are my labels ground true, that is, true classes. Test_predict is my result after prediction.

[Aneeba NAJEEB]

How to plot when we have 6 classes?

[Gledson Melotti]

Hi, You make one against all.

[Roozbeh Kh]

I have 12 classes , how to make it one agaist all 12 ?

[Peter]

Please see the Plot ROC Curve for Classification Tree example in the perfcurve discription for how to do this.

[Jhalak Mehta]

How do I get the probability?

[Hiren Mewada]

classNames = net.Layers(end).Classes;

rocSmallNet = rocmetrics(imdsTest.Labels,score,classNames);

p = plot(rocSmallNet,ShowModelOperatingPoint=false)

Answer 2

[Salma Hassan]

sir did you find the solution i have the same problem

[Gledson Melotti]

Not. If you find it, please send it to me.

[Nazia Hameed]

did u find any solution?

[Gledson Melotti]

Hello.

[predictedLabels,scores]=classify(myNet,testeImages);

cgt = double(testeImagesLabels);

cscores = scores;

figure(1)

[X,Y,T,AUC,OPTROCPT,SUBY,SUBYNAMES] = perfcurve(cgt,cscores(:,1),1);

plot(X,Y);

grid

xlabel('False positive rate')

ylabel('True positive rate')

title('ROC for Classification CNN')

[Salma Hassan]

sir i change my code to yours and i got this figure

and if i change the line into score(:,2),1 i got this

 which one is true

[Gledson Melotti]

The second figure is True.

[Win Sheng Liew]

Sir, may i have your code plss.

[Gledson Melotti]

cgt = double(testeImagesLabels); clabel = double(Test_predict); cscores = double(Probability);

figure(2) [X,Y,T,AUC,OPTROCPT,SUBY,SUBYNAMES] = perfcurve(cgt,cscores(:,1),1); plot(X,Y,'k');

[mustafa kanaan]

Please can you help me in the section, becuase I have error thanks

Answer 3

[Hiren Mewada]

[predictions,score] = classify(net, imdsTest); % To get prediction score from last layer for each class

classNames = net.Layers(end).Classes;

rocSmallNet = rocmetrics(imdsTest.Labels,score,classNames);

p = plot(rocSmallNet,ShowModelOperatingPoint=false)