The IMageNEt paper Image Net. presents the Average ROC curve for the 16 classes in the imagenet data, visit image figure. 8 in the paper. what is the known function to compute this ROC plot. As ROC plot is for a binary classification problem. Is this average ROC plot made in the picture average of ROC's for all the 16 categories. Any Help would be appreciated.


I would guess it is using the one vs all approach. You measure the auc of the ROC on each class adding the probabilities of all other classes to create the other class. Then you average among classes the obtained areas under the curve.

| improve this answer | |
  • $\begingroup$ Is it kind of micro-average given in the link scikit-learn.org/stable/auto_examples/model_selection/… $\endgroup$ – Boris May 31 '18 at 18:20
  • $\begingroup$ For example taking only one class into account, and taking its probabilities to plot ROC curve. In the end averaging all the ROC plots to form a ROC for the whole.I am trying to do it for a image classification problem but the ROC(AUC) that I am getting is almost similar for all the classifiers. @DavidMasip $\endgroup$ – Boris May 31 '18 at 18:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.