Using similarity scores from a small gallery to estimate recognition performance for larger galleries

Authors:
Amos Y. Johnson;Jie Sun;Aaron F. Bobick
Affiliations:
-;-;-
Venue:
AMFG '03 Proceedings of the IEEE International Workshop on Analysis and Modeling of Faces and Gestures
Year:
2003

Citing 1
Cited 7

Predicting large population data cumulative match characteristic performance from small population data

AVBPA'03 Proceedings of the 4th international conference on Audio- and video-based biometric person authentication

Predicting fingerprint biometrics performance from a small gallery

Pattern Recognition Letters
Modeling and Predicting Face Recognition System Performance Based on Analysis of Similarity Scores

IEEE Transactions on Pattern Analysis and Machine Intelligence
Models of large population recognition performance

CVPR'04 Proceedings of the 2004 IEEE computer society conference on Computer vision and pattern recognition
False alarm rate: a critical performance measure for face recognition

FGR' 04 Proceedings of the Sixth IEEE international conference on Automatic face and gesture recognition
An integrated prediction model for biometrics

AVBPA'05 Proceedings of the 5th international conference on Audio- and Video-Based Biometric Person Authentication
Performance evaluation and prediction for 3d ear recognition

AVBPA'05 Proceedings of the 5th international conference on Audio- and Video-Based Biometric Person Authentication
Learning small gallery size for prediction of recognition performance on large populations

Pattern Recognition

Quantified Score

Hi-index	0.00

Visualization

Abstract

We present a method to estimate recognition performance for largegalleries of individuals using data from a significantly smallergallery. This is achieved by mathematically modelling a cumulativematch characteristic (CMC) curve. The similarity scores of thesmaller gallery are used to estimate the parameters of the model.After the parameters areestimated, the rank 1 point of the modelledCMC curve isused as our measure of recognition performance. Therank 1 point (i.e.; nearest-neighbor) represents the probability ofcorrectly identifying an individual from a gallery of a particularsize; however, as gallery size increases, the rank 1 performancedecays. Our model, without making any assumptions about the gallerydistribution, replicates this effect, and allows us to estimaterecognition performance as gallery size increases without needingto physically add more individuals to the gallery. This model isevaluated onface recognition techniques using a set of faces fromthe FERET database.