Appearance-based person re-identification is particularly difficult due to varying lighting conditions and pose variations across camera views. Taking inspiration from image retrieval, in which windowed searching over locations is proven to be more effective, the authors first perform dense local feature matching using graph cuts to properly deal with the pose variation problem. However, the re-identification problem suffers from far more overlap between feature distributions. In a re-identification problem, many samples cropped from surveillance videos are heavily contaminated by external factors or internal mechanical noises, making the images from the same pedestrian totally different. These overly difficult samples would significantly degenerate the training performance. To address this problem, a query-level loss function for ranking is proposed, benefiting from taking into account the training data every query set to decrease the punishment for those morbid samples. The authors further develop a coarse-to-fine iterative algorithm, where the update in each iteration is computed by solving a gradient-based optimisation and update iteration is to refine the training data by adjusting an ‘Expected rank’ parameter. The authors present experiments to demonstrate the performance gain of the proposed method over existing template matching and ranking models.

References

1. 1)
  - 16. Farenzena, M., Bazzani, L., Perina, A., et al: ‘Person re-identication by symmetry-driven accumulation of local features’. CVPR, 2010.
2. 2)
  - 37. Zheng, W.-S., Gong, S., Xiang, T.: ‘Associating groups of people’, BMVC, 2009, (1), pp. 23.1–23.11.
3. 3)
  - 6. Wang, X., Doretto, G., Sebastian, T., et al: ‘Shape and appearance context modeling’. ICCV, 2007, pp. 1–8.
4. 4)
  - 3. Prosser, B., Zheng, W.-S., Gong, S., et al: ‘Person re-identification by support vector ranking’. BMVC, 2010, pp. 21.1–21.11.
5. 5)
  - 13. Zheng, W.-S., Gong, S., Xiang, T.: ‘Re-identification by relative distance comparison’, PAMI, 2012, 35, (3).
6. 6)
  - 33. Qin, T., Zhang, X.-D., Tsai, M.-F., et al: ‘Query-level loss functions for information retrieval’, Inf. Process. Manag., 2008, 44, (2), pp. 838–855.
7. 7)
  - 22. Kostinger, M., Hirzer, M.: ‘Large scale metric learning from equivalence constraints’. CVPR, 2012.
8. 8)
  - 21. Ba¸k, S., Charpiat, G., Corvée, E., et al: ‘Learning to match appearances by correlations in a covariance metric space’. ECCV, 2012.
9. 9)
  - 17. Gheissari, N., Sebastian, T.B., Tu, P.H., et al: ‘Person reidentification using spatiotemporal appearance’. CVPR, 2006.
10. 10)
  - 28. Kolmogorov, V., Zabih, R.: ‘What energy functions can be minimized via graph cuts?’, PAMI, 2004, 26, (2), pp. 147–159.
11. 11)
  - 19. Kviatkovsky, I., Adam, A., Rivlin, E.: ‘Color invariants for person reidentification’, PAMI, 2013, 35, (7), pp. 1622–1634.
12. 12)
  - 36. Hastie, T., Tibshirani, R., Friedman, J.: ‘The elements of statistical learning: data mining, inference, and prediction’ (Springer, New York, 2009).
13. 13)
  - 7. Ba¸k, S., Corvee, E., Bremond, F., et al: ‘Person re-identification using spatial covariance regions of human body parts’. AVSS. IEEE, August 2010.
14. 14)
  - 31. Zhu, S.C., Lee, T.S., Yuille, A.L.: ‘Region competition: unifying snakes, region growing, and Bayes/MDL for multiband image segmentation’, PAMI, 1995, (6), pp. 416–423.
15. 15)
  - 14. Zheng, W.-S., Gong, S.: ‘Person re-identification by probabilistic relative distance comparison’. CVPR, 2011.
16. 16)
  - 1. Li, W., Zhao, R., Wang, X.: ‘Human reidentification with transferred metric learning’. ACCV, 2012.
17. 17)
  - 9. Zhao, R., Ouyang, W., Wang, X.: ‘Unsupervised salience learning for person reidentification’. CVPR, 2013.
18. 18)
  - 10. Loy, C., Liu, C., Gong, S.: ‘Person re-identification by manifold ranking’. ICCV, 2013.
19. 19)
  - 4. Gray, D., Brennan, S., Tao, H.: ‘Evaluating appearance models for recognition, reacquisition, and tracking’. IEEE Int. Work. Performance Evaluating Tracking Surveillance, 2007.
20. 20)
  - 12. Gray, D., Tao, H.: ‘Viewpoint invariant pedestrian recognition with an ensemble of localized features’. ECCV, 2008.
21. 21)
  - 27. Yang, J., Simon, S., Yang, M.-H.: ‘Max-margin boltzmann machines for object segmentation’. CVPR, Columbus, 2014.
22. 22)
  - 30. Leclerc, Y.: ‘Constructing simple stable descriptions for image partitioning’, IJCV, 1989, 102, pp. 73–102.
23. 23)
  - 24. Hirzer, M., Roth, P.M., Bischof, H.: ‘Person re-identification by efficient impostor-based metric learning’. AVSS. IEEE, September 2012, pp. 203–208.
24. 24)
  - 2. Mignon, A., Jurie, F.: ‘PCCA: A new approach for distance learning from sparse pairwise constraints’. CVPR, 2012.
25. 25)
  - 26. Hiremath, P., Pujari, J.: ‘Content based image retrieval using color, texture and shape features’. ADCOM, December 2007, pp. 780–784.
26. 26)
  - 32. Delong, A., Osokin, A., Isack, H.N., et al: ‘Fast approximate energy minimization with label costs’, IJCV, 2011, 96, (1), pp. 1–27.
27. 27)
  - 23. Hirzer, M., Roth, P., Köstinger, M., et al: ‘Relaxed pairwise learned metric for person re-identification’. ECCV, 2012, pp. 780–793.
28. 28)
  - 8. Cheng, D.S., Cristani, M., Morego, V., et al: ‘Custom Pictorial Structures for Re-identification’. BMVC, 2011.
29. 29)
  - 18. Ma, B., Su, Y., Jurie, F.: ‘Local descriptors encoded by fisher vectors for person reidentification’. ECCV, 2012.
30. 30)
  - 5. Liu, C., Gong, S., Loy, C.C.C., et al: ‘Person re-identification: what features are important?’. ECCV Work., 2012.
31. 31)
  - 20. Zhao, R., Ouyang, W., Wang, X.: ‘Person re-identification by salience matching’. ICCV, 2013.
32. 32)
  - 29. Boykov, Y., Veksler, O., Zabih, R.: ‘Fast approximate energy minimization via graph cuts’, PAMI, 2001, 23, (11), pp. 1222–1239.
33. 33)
  - 15. Gong, S., Cristani, M., Loy, C., et al: ‘The re-identification challenge’. Person Re-Identification, 2014.
34. 34)
  - 35. Layne, R., Hospedales, T., Gong, S.: ‘Attributes-based re-identification’. ‘Person re-identification’ (Springer, London, 2014), pp. 93–117.
35. 35)
  - 25. Xiong, F., Gou, M., Camps, O., et al: ‘Person re-identification using Kernel-based Metric learning methods’. ECCV, 2014.
36. 36)
  - 11. Kuo, C.-H., Khamis, S., Shet, V.: ‘Person re-identification using semantic color names and RankBoost’. WACV, January 2013.
37. 37)
  - 34. Veloso, A., Gonçalves, M.: ‘Learning to rank using query-level rules’, J. Inf. Data Manag., 2010, 1, pp. 567–581.

Pedestrian re-identification via coarse-to-fine ranking

References

Related content