Improved Re-ID
Re-identification accuracy has been significantly improved by employing 3D scene information and a motion of the target.
For details see paper: Improving Person Re-identification by Viewpoint Cues, AVSS 2014

COSMATI signature
Different regions of an object are characterized by different kinds of features. Learning is performed in a covariance metric space using an entropy-driven criterion
For details see paper: Learning to Match Appearances by Correlations in a Covariance Metric Space, ECCV 2012

MRCG signature
Compact and robust representation of multiple images is obtained by using a dense grid of mean Riemannian features.
For details see paper: Multiple-shot human re-identification by mean Riemannian covariance grid, AVSS 2011

LCP signature
One-against-all learning scheme is applied to extract a discriminative representation of the human appearance. The learning is performed on a Riemannian manifold.
For details see paper: Boosted human re-identification using Riemannian manifolds, IMAVIS 2011

SCR signature
Covariance matrix descriptor together with a new spatial pyramid scheme is used to characterize the human body parts detected by histogram of oriented gradients (HOG).
For details see paper: Person re-identification using spatial covariance regions of human body parts, AVSS 2010

HAAR signature
One-against-all learning scheme is applied to extract a discriminative representation of the human appearance.
For details see paper: Person re-identification using haar-based and dcd-based signature, AVSS 2010

DCD signature
Dominant Color Descriptor together with an asymmetry-driven human body separation is applied for producing invariant human signatures.
For details see paper: Person re-identification using haar-based and dcd-based signature, AVSS 2010