SBSGAN: Suppression of Inter-Domain Background Shift for Person Re-Identification

Yan Huang,Qiang Wu,JingSong Xu,Yi Zhong

DOI: https://doi.org/10.48550/arXiv.1908.09086

2019-08-24

Abstract:Cross-domain person re-identification (re-ID) is challenging due to the bias between training and testing domains. We observe that if backgrounds in the training and testing datasets are very different, it dramatically introduces difficulties to extract robust pedestrian features, and thus compromises the cross-domain person re-ID performance. In this paper, we formulate such problems as a background shift problem. A Suppression of Background Shift Generative Adversarial Network (SBSGAN) is proposed to generate images with suppressed backgrounds. Unlike simply removing backgrounds using binary masks, SBSGAN allows the generator to decide whether pixels should be preserved or suppressed to reduce segmentation errors caused by noisy foreground masks. Additionally, we take ID-related cues, such as vehicles and companions into consideration. With high-quality generated images, a Densely Associated 2-Stream (DA-2S) network is introduced with Inter Stream Densely Connection (ISDC) modules to strengthen the complementarity of the generated data and ID-related cues. The experiments show that the proposed method achieves competitive performance on three re-ID datasets, ie., Market-1501, DukeMTMC-reID, and CUHK03, under the cross-domain person re-ID scenario.

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

This paper attempts to solve the performance degradation problem in cross - domain person re - identification (re - ID) caused by background differences. Specifically, when the training data and the test data come from different scenarios, the significant differences in the background will introduce difficulties in extracting robust pedestrian features, thus affecting the performance of cross - domain person re - identification. The paper defines this problem as the background shift problem and proposes a new generative adversarial network - Suppression of Background Shift Generative Adversarial Network (SBSGAN) to generate images with moderately suppressed backgrounds in order to reduce the background differences across domains.

SBSGAN can not only generate images with moderately suppressed backgrounds, but also maintain the integrity of foreground information, avoiding the possible damage to foreground information when directly using binary masks to remove the background. In addition, the paper also proposes a Densely Associated 2 - Stream Network (DA - 2S). By introducing the Inter Stream Densely Connection (ISDC), it strengthens the complementarity between the two input streams and further utilizes useful cues in the background (such as companions and vehicles) to enhance the representation of pedestrians.

Through experiments on three commonly used person re - identification datasets (Market - 1501, DukeMTMC - reID, and CUHK03), the paper verifies the effectiveness of the proposed method and shows its competitiveness in cross - domain person re - identification tasks.