Abstract:Multi-shot person re-identification (MsP-RID) utilizes multiple images from the same person to facilitate identification. Considering the fact that motion information may not be discriminative nor reliable enough for MsP-RID, this paper is focused on handling the large variations in the visual appearances through learning discriminative visual metrics for identification. Existing metric learning-based methods usually exploit pair-wise or triple-wise similarity constraints, that generally demands intensive optimization in metric learning, or leads to degraded performances by using sub-optimal solutions. In addition, as the training data are significantly imbalanced, the learning can be largely dominated by the negative pairs and thus produces unstable and non-discriminative results. In this paper, we propose a novel type of similarity constraint. It assigns the sample points to a set of reference points to produce a linear number of reference constraints. Several optimal transport-based schemes for reference constraint generation are proposed and studied. Based on those constraints, by utilizing a typical regressive metric learning model, the closed-form solution of the learned metric can be easily obtained. Extensive experiments and comparative studies on several public MsP-RID benchmarks have validated the effectiveness of our method and its significant superiority over the state-of-the-art MsP-RID methods in terms of both identification accuracy and running speed.

Easy Identification from Better Constraints: Multi-Shot Person Re-Identification from Reference Constraints.

Adaptive Metric Learning For People Re-Identification

People Re-Identification with Local Distance Comparison Using Learned Metric.

Integrated Metric Learning with Adaptive Constraints for Person Re-Identification

Adaptive Metric Learning in Local Distance Comparison for People Re-identification

Multiple-Shot People Re-Identification by Patch-Wise Learning.

Multiple-shot Person Re-Identification Via Fair Set-Collaboration Metric Learning

A Multi-Constraint Similarity Learning with Adaptive Weighting for Visible-Thermal Person Re-Identification

Large Margin Relative Distance Learning for Person Re-Identification

Person Re-Identification Based On Set-To-Set Metric Learning

Person Re-identification Using Data-Driven Metric Adaptation.

Person Re-Identification by Semantic Region Representation and Topology Constraint

Person Re-Identification by Pose Invariant Deep Metric Learning with Improved Triplet Loss.

A Subspace Learning Approach to Multishot Person Reidentification

Metric Learning by Simultaneously Learning Linear Transformation Matrix and Weight Matrix for Person Re-Identification

Person reidentification via structural deep metric learning

Equidistance Constrained Metric Learning for Person Re-Identification

Person Re-Identification Via Multiple Coarse-To-Fine Deep Metrics

Person Re-Identification by Dual-Regularized KISS Metric Learning.

Person Re-Identification by Discriminant Analytical Least Squares Metric Learning

Person Re-Identification by Integrating Metric Learning and Support Vector Machine