In the field of demographic attribute classification, race estimation is perhaps the least studied topic in the literature. CNN-based approaches report the best results to the day, but they are computational expensive for practical applications. We propose a simpler approach by combining local appearance and geometrical features to describe face images, and to exploit the race information from different face parts by means of a component-based methodology. Experimental results obtained in the FERET subset from EGA database, with traditional but effective classifiers like Random Forest and Support Vector Machines, are very close to those achieved with a recent deep learning proposal.
On Combining Face Local Appearance and Geometrical Features for Race Classification / Becerra-Riera, Fabiola; Méndez Llanes, Nelson; Morales-González, Annette; MENDEZ VAZQUEZ, Heydi; Tistarelli, Massimo. - Lecture Notes in Computer Science 11401:(2018), pp. 567-574. (Intervento presentato al convegno Iberoamerican Congress on Pattern Recognition - CIARP 2018) [10.1007/978-3-030-13469-3_66].
On Combining Face Local Appearance and Geometrical Features for Race Classification
Heydi Méndez-Vázquez;Massimo Tistarelli
2018-01-01
Abstract
In the field of demographic attribute classification, race estimation is perhaps the least studied topic in the literature. CNN-based approaches report the best results to the day, but they are computational expensive for practical applications. We propose a simpler approach by combining local appearance and geometrical features to describe face images, and to exploit the race information from different face parts by means of a component-based methodology. Experimental results obtained in the FERET subset from EGA database, with traditional but effective classifiers like Random Forest and Support Vector Machines, are very close to those achieved with a recent deep learning proposal.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
