Learning local features for age estimation on real-life faces

  • Authors:
  • Caifeng Shan

  • Affiliations:
  • Philips Research, Eindhoven, Netherlands

  • Venue:
  • Proceedings of the 1st ACM international workshop on Multimodal pervasive video analysis
  • Year:
  • 2010

Quantified Score

Hi-index 0.00

Visualization

Abstract

In this paper, we investigate age estimation on real-life faces acquired in unconstrained conditions. This is a challenging but relatively understudied problem, with interesting applications in many areas (e.g., visual surveillance). We use the large dataset recently collected in (Gallagher and Chen, 2009). Appearance features, including Local Binary Patterns (LBP) and Gabor features are exploited as face representation. We adopt Adaboost to learn the discriminative local features. More specifically, we propose to learn the discriminative LBP-Histogram bins for face age estimation. Our experiments illustrate Support Vector Machines (SVM) based on the boosted features outperform the methods in (Gallagher and Chen 2009), achieving reasonable performance given the difficulty of the dataset.