ROC Curve for Multiclass RCNN

I have created a multiclass RCNN via the trainRCNNObjectDetector function and have created a results structure of a validation set of data, which is made up of 3 fields: Bounding Boxes, Scores and Labels, with their being two potential labels.

In reading the plotroc and perfcurve documentation, I haven't been able to find any method for obtaining performance or ROC curves for this multiclass problem. Is there any current methodology to attack this?

[Don Zheng]

The example given in 'plotroc' deals with four classes and each output have scores for four classes. Can you clarify what you mean by "multiclass problem" and why 'plotroc' does not work?

[Timothy Kain]

The trainRCNNObjectDetector function in MATLAB does not train an SVM classifier for each object class, without an SVM classifier I am unable to extract the values of TPR and FPR that would be needed to form a ROC curve.

[zainab haddad]

hi sir, please can you help me by your multiclass RCNN if i don't bothred you

[neora asulin]

HI

do you have a answer how to training multi class with rcnn ?