Riemannian Set-level Common-Near-Neighbor Analysis for Multiple-shot Person Re-identification

Multiple-shot person re-identification deals with the problem to build the correspondence between the im-ages of the same person appearing at different sites captured by onsite deployed cameras. The difficul-ty stems from large within-class but small between-class variations caused by the change of person ap-pearance and environment. Traditional methods on feature/signature design and/or distance/dissimilarity exploration have been largely investigated, resulting in a quick slowing down of the performance improvemen-t. This paper proposes a novel solution called “Rie-mannian Set-level Common-Near-Neighbor Analysis” by absorbing the essence of two distinctive and effec-tive state-of-the-art models. More concretely, it gener-ates the discriminative covariance-based representation for each set of images following the Mean Riemanni-an Covariance Grid approach, while at the same time creatively realizes the set-level neighborhood informa-tion based ranking inheriting the key idea of sample-level Common-Near-Neighbor Analysis. Experiments have been conducted on widely-used benchmark dataset-s, showing significant performance improvement over state-of-the-art methods.