Cross-View GAN Based Vehicle Generation for Re-identification

Yi Zhou and Ling Shao

Abstract

Automatic vehicle re-identification (re-ID) is highly valuable and significant in public transportation systems, but has not achieved much progress since the visual appearances vary hugely across different viewpoints of a vehicle. Feature matching in this problem is extremely difficult, and traditional person re-ID algorithms cannot be suitably applied to vehicles. However, image generation by convolutional generative adversarial networks (GANs), which has obtained breakthrough progress, inspires us to generate vehicles in different viewpoints from only one visible view to tackle vehicle re-ID. In this work, we propose a new deep architecture, called Cross-View Generative Adversarial Network (XVGAN), to learn the features of vehicle images captured by cameras with disjoint views, and take the features as conditional variables to effectively infer cross-view images. Finally, the features of the original images are combined with the features of generated images in other views to learn distance metrics for vehicle re-ID.

Session

Orals - Matching

Files

PDF iconPaper (PDF)

DOI

10.5244/C.31.186
https://dx.doi.org/10.5244/C.31.186

Citation

Yi Zhou and Ling Shao. Cross-View GAN Based Vehicle Generation for Re-identification. In T.K. Kim, S. Zafeiriou, G. Brostow and K. Mikolajczyk, editors, Proceedings of the British Machine Vision Conference (BMVC), pages 186.1-186.12. BMVA Press, September 2017.

Bibtex

            @inproceedings{BMVC2017_186,
                title={Cross-View GAN Based Vehicle Generation for Re-identification},
                author={Yi Zhou and Ling Shao},
                year={2017},
                month={September},
                pages={186.1-186.12},
                articleno={186},
                numpages={12},
                booktitle={Proceedings of the British Machine Vision Conference (BMVC)},
                publisher={BMVA Press},
                editor={Tae-Kyun Kim, Stefanos Zafeiriou, Gabriel Brostow and Krystian Mikolajczyk},
                doi={10.5244/C.31.186},
                isbn={1-901725-60-X},
                url={https://dx.doi.org/10.5244/C.31.186}
            }