How Matlab measure accuracy, sensitivity and specificity in faster rcnn?

Hi, I saw from the post using confusion matrix to measure accuracy, sensitivity and specificity. however, in faster rcnn, there are scores and w,h,x, and y, how am i going to measure these accuracy, sensitivity and specificity in faster rcnn? in detectorPrecisionRecall.m, there only have true positive and false positive, is it i need to remodify these code myself? Am i still need to use confusion matrix to do?

Answer 1

[Shashank Gupta]

Check out evaluateDetectionPrecision. which is same as you mentioned about detectorPrecisionRecall. you must have seen that this function are widely used in the object detector examples, its because the output you get from these functions are "precision" and "recall", the two important performance metric which is nothing but sensitivity and Positive predictive value. You can use these metric to calculate almost all important performance criteria. Thus you may not need to use confusion matrix at all.

I hope this clear your confusion.

Cheers.

[winnie poh]

Thank you for the explanation