This paper presents a study on personal aesthetics, a recentsoft biometrics application where the goal is to recognize people by con-sidering the images they like. Here we propose a multi-level approach,where each level is intended as a low-dimensional space where the im-ages preferred by a user can be projected, and similar images are mappednearby, namely a Counting Grid. Multiple levels are generated by adopt-ing Counting Grids at different resolutions, corresponding to analyzeimages at different grains. Each level is then associated to an exemplarSupport Vector Machine, which separates the images of an individualfrom the rest of the users. Putting together multiple levels gives a bat-tery of classifiers whose performances are very good: on a dataset of 200users, and 40K images, using 5 preferred images as biometric templategives 97% of probability of guessing the correct user; as for the verifica-tion capability, the equal error rate is 0.11. The approach has also beentested with diverse comparative methods and different features, showingthat color image properties are crucial to encode the personal aesthetics,and that high-level information (as the objects within the images) couldbe very effective, but current methods are not robust enough to catch it

Recognizing People by Their Personal Aesthetics: A Statistical Multi-level Approach

Segalin, Cristina;PERINA, Alessandro;CRISTANI, Marco
2014

Abstract

This paper presents a study on personal aesthetics, a recentsoft biometrics application where the goal is to recognize people by con-sidering the images they like. Here we propose a multi-level approach,where each level is intended as a low-dimensional space where the im-ages preferred by a user can be projected, and similar images are mappednearby, namely a Counting Grid. Multiple levels are generated by adopt-ing Counting Grids at different resolutions, corresponding to analyzeimages at different grains. Each level is then associated to an exemplarSupport Vector Machine, which separates the images of an individualfrom the rest of the users. Putting together multiple levels gives a bat-tery of classifiers whose performances are very good: on a dataset of 200users, and 40K images, using 5 preferred images as biometric templategives 97% of probability of guessing the correct user; as for the verifica-tion capability, the equal error rate is 0.11. The approach has also beentested with diverse comparative methods and different features, showingthat color image properties are crucial to encode the personal aesthetics,and that high-level information (as the objects within the images) couldbe very effective, but current methods are not robust enough to catch it
image processing
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: http://hdl.handle.net/11562/878985
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact