Scalable Person Re-identification on Supervised Smoothed Manifold

By Song Bai, Xiang Bai, Qi Tian. Computer Vision and Pattern Recognition, 2017

339 citations recorded by the source corpus.

Most existing person re-identification algorithms either extract robust visual features or learn discriminative metrics for person images. However, the underlying manifold which those images reside on is rarely investigated. That arises a problem that the learned metric is not smooth with respect to the local geometry structure of the data manifold. In this paper, we study person re-identification with manifold-based affinity learning, which did not receive enough attention from this area. An unconventional manifold-preserving algorithm is proposed, which can 1) make best use of supervision from training data, whose label information is given as pairwise constraints, 2) scale up to large repositories with low on-line time complexity, and 3) be plunged into most existing algorithms, serving as a generic postprocessing procedure to further boost the identification accuracies. Extensive experimental results on five popular person re-identification benchmarks consistently demonstrate the effectiveness of our method. Especially, on the largest CUHK03 and Market-1501, our method outperforms the state-of-the-art alternatives by a large margin with high efficiency, which is more appropriate for practical applications.

Subfields: Biomedical Engineering, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/be3d6508739d8fac084cd829cb9d93b507f06b97

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Biomedical Engineering

1. Liang Zheng
2. Yi Yang
3. Shiliang Zhang
4. Qi Tian
5. Mang Ye
6. Wen Gao
7. Pong Chi Yuen
8. Tao A. Xiang

In Computer Vision and Pattern Recognition

1. Yi Yang
2. Luc Van Gool
3. Dacheng Tao
4. Liang Zheng
5. Xiaogang Wang
6. Ling Shao
7. Qi Tian
8. Ming-Hsuan Yang

Citations in the corpus

• FastReID: A Pytorch Toolbox for General Instance Re-identification (2020, methodology): “The state-of-the-art algorithms published in CVPR, ICCV, ECCV during 2018-2020 are listed in Table 1, FastReID achieves the best performance on Market1501 96.3%(90.3%), DukeMTMC 92.4%(83.2”
• Bilinear Attention Networks for Person Retrieval (2019, background): “Structures, such as orthogonality constraints [37] and geometry contraints [1], have also shown to help achieving better, or more robust, decisions in person re-identification.”
• Camera Style Adaptation for Person Re-identification (2017, neutral): “where ∈ [0, 1].”
• Deep Representation Learning With Part Loss for Person Re-Identification (2017, background): “47%, which is lower than some metric learning methods [5], [11], [12], [65].”
• Robust Anchor Embedding for Unsupervised Video Person re-IDentification in the Wild (2018, background): “Note that the performance of feature learning process might be improved with other advanced deep learning [26,44] or re-ranking methods [1, 2, 46, 57].”
• Cross-Modality Person Re-Identification With Shared-Specific Feature Transfer (2020, methodology): “Recent self-attention [41, 29] and graph-based methods [1, 34, 35, 47] tried to model the relationship between sample pairs.”
• A Pose-Sensitive Embedding for Person Re-identification with Expanded Cross Neighborhood Re-ranking (2017, background): “We studied the impact of changing these on Market and Duke datasets and found that it is subtle in the range for t ∈ [2, 4] and q ∈ [4, 10], the performance drops between∼0.”
• Multi-level Factorisation Net for Person Re-identification (2018, neutral): “1 SSM [2] 82.”
• Mask-Guided Contrastive Attention Model for Person Re-identification (2018, neutral): “4 Scalable [1] CVPR 2017 82.”
• Person Re-identification in the Wild (2016, methodology): “In fact, re-ranking methods [2,46,51] will be critical for scalable re-ID.”
• Human Semantic Parsing for Person Re-identification (2018, neutral): “0 – – SSM [3] 68.”
• Distilled Person Re-Identification: Towards a More Scalable System (2019, background): “Fast adaptation [5, 39] is for reducing computation cost in training.”
• Multi-Task Learning with Low Rank Attribute Embedding for Multi-Camera Person Re-Identification (2018, background): “[31] investigate person re-identification task with manifold-based affinity learning and use an unconventional manifold-preserving algorithm to improve boost identification accuracy.”
• Multiple Instance Detection Network with Online Instance Classifier Refinement (2017, methodology): “The proposal labelling procedure is also related to the semi-supervised label propagation method [2, 34].”
• Channel Augmented Joint Learning for Visible-Infrared Recognition (2021, background): “Visible-Infrared Person Re-IDentification (VI-ReID) is a cross-modality person recognition problem, which aims at matching daytime visible and night-time infrared images [2, 48, 16, 57].”