Boosted human re-identification using Riemannian manifolds

Abstract

This paper presents an appearance-based model to address the human reidentication problem. Human re-identication is an important and still unsolved task in computer vision. In many systems there is a requirement to identify individuals or determine whether a given individual has already appeared somewhere in a network of cameras. The human appearance obtained in one camera is usually dierent from the ones obtained in another camera. In order to re-identify people a human signature should handle dierence in illumination, pose and camera parameters. The paper focuses on a new appearance model based on Mean Riemannian Covariance (MRC) patches extracted from tracks of a particular individual. A new similarity measure using Riemannian manifold theory is also proposed to distinguish sets of patches belonging to a specic individual. We investigate the signicance of the MRC patches based on their reliability extracted during tracking and their discriminative power obtained by a boosting scheme. The methods are evaluated and compared with the state of the art using benchmark video sequences from the ETHZ and the i-LIDS datasets. The re-identication performance is presented using the cumulative matching characteristic (CMC) curve. We demonstrate that the proposed approach outperforms state of the art methods. Finally, the results of our approach are shown on two other more pertinent datasets.

Collaboratively Regularized Nearest Points for Set Based Recognition

“…For person re-identification, we follow [9] on using the iLIDS-MA and iLIDS-AA datasets [1] for evaluation, along with an additional dataset CAVIAR4REID [3]. These datasets have interesting and complementary properties, thus being informative for comparison.…”

Multi-view face recognition from single RGBD models of the faces

“…When there is no overlap between the fields of view of the cameras involved, person re-identification becomes a fundamental issue in multi-camera face recognition (Bąk, Corvee, Bremond, Thonnat, 2012, Bedagkar-Gala, Shah, 2014, Cai, Huang, Tan, 2008, Gong, Cristani, Loy, Hospedales, 2014, Mazzon, Tahir, Cavallaro, 2012, de Oliveira, de Souza Pio, 2009, Satta, Fumera, Roli, 2012, Zhu, Luo, Wang, Tang, 2014. The notion of reidentification addresses the following issue related to face recognition in the wild: If a group of people is being tracked by a network of non-overlapping cameras, how can we ensure that the face fragments extracted from two different cameras belong to the same individual?…”

Key Person Aided Re-identification in Partially Ordered Pedestrian Set

“…Figure 1: From the traditional pedestrian set to temporally partially ordered set multiple images of a person provide more clues to differentiate pedestrians from each other. For example, the appearance features extracted by multiple images were accumulated or averaged into a single signature [1,3]. The temporal correlation of multiple images was also exploited to build the spatial-temporal appearance representation for person re-id [11,27].…”