Person Re-Identification by Support Vector Ranking

By Bryan James Prosser, Weishi Zheng, Shaogang Gong, Tao A. Xiang. British Machine Vision Conference, 2010

772 citations recorded by the source corpus.

Solving the person re-identification problem involves matching observation s of individuals across disjoint camera views. The problem becomes particularly hard in a busy public scene as the number of possible matches is very high. This is further compounded by significant appearance changes due to varying lighting conditions, vie wing angles and body poses across camera views. To address this problem, existing approaches focus on extracting or learning discriminative features followed by template matching using a distance measure. The novelty of this work is that we reformulate the person reidentification problem as a ranking problem and learn a subspace where th e potential true match is given highest ranking rather than any direct distance measure. By doing so, we convert the person re-identification problem from an absolute scoring p roblem to a relative ranking problem. We further develop an novel Ensemble RankSVMto overcome the scalability limitation problem suffered by existing SVM-based ranking methods. This new model reduces significantly memory usage therefore is much more scalable, whilst maintaining high-level performance. We present extensive experiments to demonstrate the performance gain of the proposed ranking approach over existing template matching and classification models.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/cdf24d7eecef4cc0d4d29c142baa1c5dbbc23a83

How researchers see this paper

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

In Computer Vision and Pattern Recognition

1. Yi Yang
2. Xiaogang Wang
3. Liang Zheng
4. Luc Van Gool
5. Qi Tian
6. Wanli Ouyang
7. Dacheng Tao
8. Chunhua Shen

Citations in the corpus

• Person Re-Identification by Camera Correlation Aware Feature Augmentation (2017, neutral): “based methods [6], [8], [9], [11], [12], [13], [14], [15], [34],”
• Person Re-identification: What Features Are Important? (2012, methodology): “They can be learned either through boosting [7], rank learning [4], or distance metric learning [5].”
• Multi-Task Learning with Low Rank Attribute Embedding for Multi-Camera Person Re-Identification (2018, methodology): “RSVM [71] KISSME [11] R-CRF [36] K-CRF [36] MTL-LOREA”
• Beyond frontal faces: Improving Person Recognition using multiple cues (2015, background): “Existing work is mainly focused on metric learning [27, 17, 22], mid-level feature learning [16, 39, 40, 10, 20, 26].”
• Distilled Person Re-Identification: Towards a More Scalable System (2019, background): “(2) Ensembles of all teachers by distance fusion with weights learned by RankSVM [43] on the validation set and with equal weights.”
• Person re-identification by probabilistic relative distance comparison (2011, methodology): “We demonstrate the advantage of our approach over both the Boosting [7] and RankSVM [11] based methods through experiments.”
• Person Re-identification by Salience Matching (2013, methodology): “PRDC [23], LMNNR [4], PCCA [18], and attribute-based PRDC (aPRDC) [13], a boosting approach (ELF) [8] and Rank SVM (RankSVM) [19].”
• Cross Dataset Person Re-identification (2014, methodology): “Prosser et al.[25] formulated the person re-identification problem as a ranking problem and applied the Ensemble RankSVM to learn a subspace where the potential true match get the highest rank.”
• DeepReID: Deep Filter Pairing Neural Network for Person Re-identification (2014, methodology): “[35] use boosting and RankSVM, respectively, to select features and compute the distance between images.”
• Person Re-identification by Deep Learning Multi-scale Representations (2017, background): “Recently, deep learning re-id models [26, 4, 48, 56, 44, 77, 64, 7, 11, 65, 9, 60, 27] start to take over and have obtained impressive performance.”
• Deep Metric Learning for Person Re-identification (2014, methodology): “From the results we can see that our method outperforms DTRSVM significantly and even approach the performance of some methods in single database setting, such as ELF [12] and RDC [2].”
• Multi-Task Learning with Low Rank Attribute Embedding for Person Re-Identification (2015, methodology): “We compare our algorithm with 11 learning-based methods 2: Relaxed Pairwise Metric Learning (RPML) [17], Probabilistic Relative Distance Comparison (PRDC) [50], RankSVM (RSVM) [38], Salmatch [47], LMF [48], Pairwise Constrained Component Analysis (PCCA) [11], regularized PCCA (rPCCA) [42], Keep It S”
• Deep Attributes Driven Multi-Camera Person Re-identification (2016, background): “Compared works that learn distance metrics for person ReID include RPML [9], PRDC [16], RSVM [49], Salmatch [45], LMF [46], PCCA [8], KISSME [12], kLFDA [13], KCCA [47],TSR [48], EPKFM [18],LOMO + XQDA [19],MRankPRDC [26], MRank-RSVM [26], RQDA [50], MLAPG [22] and CSL [21].”
• Person Re-identification by Video Ranking (2014, methodology): “colour and intensity gradient histograms) using a pair of single-shot person images [11,35,20,49].”
• Unsupervised Salience Learning for Person Re-identification (2013, methodology): “Discriminative models like SVM and boosting [25, 13, 15] are widely used for feature learning.”
• ABD-Net: Attentive but Diverse Person Re-Identification (2019, background): “Many part-based methods have achieved superior performance [2, 3, 26, 27, 28, 4, 5, 29, 6, 30, 31, 8, 32].”