Localized Multi-Feature Metric Learning for Image Set Based Face Recognition - 2015
This paper presents a new approach to image set based face recognition, where each coaching and testing example could be a set of face images captured from varying poses, illuminations, expressions and resolutions. While a range of image set primarily based face recognition strategies have been proposed in recent years, most of them model each face image set as one linear subspace or because the union of linear subspaces, that could lose some discriminative info for face image set illustration. To address this shortcoming, we tend to propose exploiting statistics info as feature representations for face image sets, and develop a localized multi-kernel metric learning (LMKML) algorithm to effectively combine different statistics for recognition. Moreover, we tend to propose a localized multi-kernel multi-metric learning (LMKMML) method to jointly learn multiple feature-specific distance metrics in the kernel areas, one for every statistic feature, to raised exploit complementary data for recognition. Our strategies achieve state-of-the-art performance on four widely used video face datasets together with the Honda, Mobo, YouTube Celebrities (YTC), and YouTube Face (YTF) datasets.
Did you like this research project?
To get this research project Guidelines, Training and Code... Click Here
- Multiscale Logarithm Difference Edgemaps for Face Recognition Against Varying Lighting Conditions - 2015
- Learning Compact Binary Face Descriptor for Face Recognition - 2015