Style Normalization and Restitution for Generalizable Person Re-Identification

By Xin Jin, Cuiling Lan, Wenjun Zeng, Zhibo Chen, Li Zhang. Computer Vision and Pattern Recognition, 2020

401 citations recorded by the source corpus.

Existing fully-supervised person re-identification (ReID) methods usually suffer from poor generalization capability caused by domain gaps. The key to solving this problem lies in filtering out identity-irrelevant interference and learning domain-invariant person representations. In this paper, we aim to design a generalizable person ReID framework which trains a model on source domains yet is able to generalize/perform well on target domains. To achieve this goal, we propose a simple yet effective Style Normalization and Restitution (SNR) module. Specifically, we filter out style variations (e.g., illumination, color contrast) by Instance Normalization (IN). However, such a process inevitably removes discriminative information. We propose to distill identity-relevant feature from the removed information and restitute it to the network to ensure high discrimination. For better disentanglement, we enforce a dual causal loss constraint in SNR to encourage the separation of identity-relevant features and identity-irrelevant features. Extensive experiments demonstrate the strong generalization capability of our framework. Our models empowered by the SNR modules significantly outperform the state-of-the-art domain generalization approaches on multiple widely-used person ReID benchmarks, and also show superiority on unsupervised domain adaptation.

Subfields: Biomedical Engineering, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/9c364f90dbd09e5366c2d492cd12f989bf8588e6

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. Li Liu
2. Erwin M. Bakker
3. Ling Shao
4. Qian Du
5. Michael S. Lew
6. Wei Chen
7. Yu Liu
8. Nan Pu

In Computer Vision and Pattern Recognition

1. Ling Shao
2. Qian Du
3. Li Liu
4. Fan Zhu
5. Wei Li
6. Yi Zhou
7. Jie Qin
8. Ran Tao

Citations in the corpus

• Single-Source Domain Expansion Network for Cross-Scene Hyperspectral Image Classification (2022, background): “high generalization and high resolution of the network [17].”
• Domain Generalization: A Survey (2021, neutral): “In person re-identification (re-ID), domain shift mainly affects image contrast and style, and therefore, can be mitigated by using IN-equipped CNNs [59], [188].”
• Normalization Techniques in Training DNNs: Methodology, Analysis and Application (2020, methodology): “Furthermore, the idea of BN in learning invariant representation is used for person re-identification (ReID) [247], [248], [249] and stereo match [250], [251], where the data from different cameras can be viewed as different ‘domains’.”
• Learning to Generate Novel Domains for Domain Generalization (2020, neutral): “We evaluate L2A-OT on three homogeneous DG benchmark datasets including digit recognition [24,12,38], PACS [25] and Office-Home [51] and a heterogeneous DG task in the form of cross-domain person re-identification (re-ID) [58,59,32,61,22].”
• Lifelong Person Re-Identification via Adaptive Knowledge Accumulation (2021, background): “While advanced deep learning methods [55, 49, 30, 46, 38, 31, 47] have shown powerful abilities for ReID [35, 10], their training process is limited heavily by a fixed and stationary dataset [52, 54, 40].”