110 likes | 241 Views
Algorithms and Applications in Computer Vision. Lihi Zelnik -Manor lihi@ee.technion.ac.il ROC. Precision-Recall. circles. Not circles. Truth:. Not circles. Result:. circles. Precision-Recall. circles. Not circles. Truth:. Not circles. FN. Result:. circles. TP. FP. TN.
E N D
Algorithms and Applications in Computer Vision LihiZelnik-Manor lihi@ee.technion.ac.il ROC
Precision-Recall circles Not circles Truth: Not circles Result: circles
Precision-Recall circles Not circles Truth: Not circles FN Result: circles TP FP TN
Confusion matrix P = Positive N = Negative
ROC curve AUC = Area Under Curve
F-measure The harmonic mean (mean of rates) of the precision and recall
True/false positives The distance threshold affects performance True positives = # of detected matches that are correct False positives = # of detected matches that are incorrect 50 true match 75 200 false match feature distance
# true positives # matching features (positives) # false positives # unmatched features (negatives) Evaluating the results How can we measure the performance of a feature matcher? 1 0.7 truepositiverate 0 1 false positive rate 0.1
Evaluating the results How can we measure the performance of a feature matcher? # true positives # matching features (positives) # false positives # unmatched features (negatives) ROC curve (“Receiver Operator Characteristic”) 1 0.7 truepositiverate 0 1 false positive rate 0.1 • ROC Curves • Generated by counting # correct/incorrect matches, for different threholds • Want to maximize area under the curve (AUC) • Useful for comparing different feature matching methods • For more info: http://en.wikipedia.org/wiki/Receiver_operating_characteristic