Notwithstanding the enhancement obtained in the last decade researches, the recognition of facial attributes is still today a trend. Besides the mere face recognition, the singular face features, like mouth, nose and hair, are considered as soft biometrics; these can be useful for human identification in cases the face is partially occluded, and only some regions are visible. In this paper we propose a model generated by transfer learning approach for the recognition of the face attributes. Also, an unsupervised clustering model is described, which is in charge of dividing and grouping faces based on their characteristics. Furthermore, we show how clusters can be evaluated by a compact summary of them, and how Deep Learning models should be properly trained for attribute prediction tasks.

Transfer Learning for Facial Attributes Prediction and Clustering

Barra S.
;
2019-01-01

Abstract

Notwithstanding the enhancement obtained in the last decade researches, the recognition of facial attributes is still today a trend. Besides the mere face recognition, the singular face features, like mouth, nose and hair, are considered as soft biometrics; these can be useful for human identification in cases the face is partially occluded, and only some regions are visible. In this paper we propose a model generated by transfer learning approach for the recognition of the face attributes. Also, an unsupervised clustering model is described, which is in charge of dividing and grouping faces based on their characteristics. Furthermore, we show how clusters can be evaluated by a compact summary of them, and how Deep Learning models should be properly trained for attribute prediction tasks.
2019
978-981-15-1300-8
978-981-15-1301-5
Attribute clustering; Cluster summary; Face attributes; k-means; Transfer learning
File in questo prodotto:
File Dimensione Formato  
Transfer_Learning_for_Facial_AttributesPrediction_and_Clustering___DependSys2019___LNCS.pdf

Solo gestori archivio

Tipologia: versione pre-print
Dimensione 1.79 MB
Formato Adobe PDF
1.79 MB Adobe PDF   Visualizza/Apri   Richiedi una copia

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: https://hdl.handle.net/11584/288719
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 19
  • ???jsp.display-item.citation.isi??? ND
social impact