Eg: Among 10 faces suppose 8 are correctly identified and 2 are false identification
recognition rate= (no. of correctly identified images / Total no. of images)*100
=(8/10)*100
=80
yes the formula is correct. Here, images are face images, which are correctly identified. The identification is done manually, for known answers. This is for the purpose of testing the system accuracy. We must know the answer before hand. Only, then we can test a system
80% and false rate is 20%. But you need thousands of different faces To reach credible ratés.
As Kaya said you consider measuring the false acceptance rate (FAR) and the False Rejection at different thresholds. then you also plot the ROC curve to see the performance of the system
I agree with Kaya and Azom. ROC curve with good values for your proposed work finds place in a good journal
I think False positives , True positive, false negative FAR-,is calculated for Face Detection not for Face Recognition. Then how to calculate for Face Recognition .I have drawn ROC curve for Face Detection, Is there separate formula for Face Recognition
An introduction to ROC analysis - http://people.inf.elte.hu/kiss/13dwhdm/roc.pdf.explained detail about ROC
try this http://citeseerx.ist.psu.edu/viewdoc/download;jsessionid=AC6D7020A1F071376262AEDCC2223B87?doi=10.1.1.115.9985&rep=rep1&type=pdf
- Badges
- Science topic
- Similar topics
- Geoscience
- Asia
- India