Rethinking the Distribution Gap of Person Re-identification with Camera-Based Batch Normalization

Zhuang, Zijie; Wei, Longhui; Xie, Lingxi; TianYu, Zhang; Zhang, Hengheng; Wu, Haozhe; Ai, Haizhou; Tian, Qi

doi:10.1007/978-3-030-58610-2_9

Cited by 139 publications

(48 citation statements)

References 52 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…On the CUHK03 dataset, we compare 12 methods with our solution, and the results are shown in Table 3. These various state-of-the-art methods include local feature models [7,57,60,63], attention mechanism models [35,36,55], multibranch networks [8,25,30], multiscale networks [56,62,64], and other deep models [45,49,53,54,58,61]. It T A B L E 1 Statistics of Market-1501, DukeMTMC-reID, and CUHK03 datasets, wherein N id , N im , and N cam represent the number of the identity, image, and camera, respectively.…”

Section: Comparison With State-of-the-art Methodsmentioning

confidence: 99%

Multi‐branch network with hierarchical bilinear pooling for person reidentification

et al. 2021

View full text Add to dashboard Cite

Because of issues such as viewpoint changes, posture variations, and background cluttering, the task of person reidentification (Re-ID) remains challenging. The model of combining global features and part features has been widely used in person Re-ID technology in recent years, but these efforts ignored feature interaction between the convolutional layers and thus lost detailed information conducive to identifying pedestrians under different cameras. To achieve interaction between hierarchical features, a multibranch network with hierarchical bilinear pooling (MBN-HBP) is proposed. The network consists of a global branch, a part-level branch, and a hierarchical bilinear pooling (HBP) branch. The person features extracted by the network include not only global and part-level features but also detailed HBP features. The final feature representation will be more robust to deal with the complex surveillance environment. By conducting comprehensive experiments, competitive performance on the Market-1501, DukeMTMC-Re-ID, and CUHK03 datasets is obtained.This is an open access article under the terms of the Creative Commons Attribution-NonCommercial License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes.

show abstract

Section: Comparison With State-of-the-art Methodsmentioning

confidence: 99%

Multi‐branch network with hierarchical bilinear pooling for person reidentification

et al. 2021

View full text Add to dashboard Cite

show abstract

“…It iteratively clusters images by style features extracted from middle layers of the networks, and utilizes clustering labels to guide the cluster-based BatchNorm for distribution alignment. This normalization method is first proposed to reduce the distribution gap between different cameras by standardizing inputs according to the camera-related statistics [11]. However, our framework utilizes clustering labels instead of camera annotations for the Batch Normalization.…”

Section: Proposed Methodsmentioning

confidence: 99%

“…To overcome this problem, Jin et al [10] proposed Style Normalization and Restitution (SNR) module which compensates normalized features with identity-relevant features that is distilled from the residual information. As emphasized in [11], it is of great importance to bridge the distribution gap between all cameras. Therefore, Zhuang et al [11] proposed CBN to eliminate distribution inconsistency between all cameras, which makes model more generalizable by preventing them from overfitting to any camera.…”

Section: Related Workmentioning

confidence: 99%

“…As emphasized in [11], it is of great importance to bridge the distribution gap between all cameras. Therefore, Zhuang et al [11] proposed CBN to eliminate distribution inconsistency between all cameras, which makes model more generalizable by preventing them from overfitting to any camera.…”

Section: Related Workmentioning

confidence: 99%

“…Thus, exploring IN layers to enhance the generalization capability of ReID models becomes popular [9,10]. From the perspective of normalization technology, Zhuang et al [11] proposed Camera-based Batch Normalization (CBN) to eliminate the distribution inconsistency across all cameras in ReID dataset for better generalization capability. The great performance of CBN largely depends on camera labels to bridge the distribution gap among different camera images.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Cluster-Based Distribution Alignment For Generalizable Person Re-Identification

Zhu

Chang

Xiao

et al. 2021

2021 IEEE International Conference on Multimedia &Amp; Expo Workshops (ICMEW)

View full text Add to dashboard Cite

Person re-identification (ReID) aims to identify a person of interest across surveillance cameras. Although supervised ReID methods have made a great success, most of them suffer a large drop in performance on unseen domains. The key idea to solve this problem lies in the distribution alignment of pedestrian images. In this paper, we design a generalizable person ReID framework which is trained on the source domain but still generalizes well on unseen domains. To achieve this goal, our framework adaptively searches and aligns the underlying distributions among images through cluster-based Batch Normalization (BatchNorm). Specifically, the framework iteratively clusters images based on the style features extracted from middle layers of the networks. The clustering labels are then utilized to guide the cluster-based BatchNorm, which standardizes features in the same cluster within a batch. Extensive experiments on several ReID benchmarks demonstrate the effectiveness of our approach.

show abstract

Learning to generalize aerial person re‐identification using the meta‐transfer method

Peng

Lü

et al. 2023

Concurrency and Computation

View full text Add to dashboard Cite

Person re-identification (Re-ID) aims to retrieve a person of interest across multiple nonoverlapping cameras. In recent years, to enable person Re-ID technology to play out its application value in real-world scenarios, visual surveillance through unmanned aerial vehicle (UAV) platforms has received intense attention, and aerial person datasets have been constructed. However, the pedestrian images captured by ground cameras and those captured by UAVs exhibit great differences. Person Re-ID methods based on ground person images have difficulty performing Re-ID on aerial person images. In this paper, we first use a meta-transfer method to learn to gener-aerial imagery, meta-transfer learning, person re-identification

show abstract

Rethinking the Distribution Gap of Person Re-identification with Camera-Based Batch Normalization

Cited by 139 publications

References 52 publications

Multi‐branch network with hierarchical bilinear pooling for person reidentification

Multi‐branch network with hierarchical bilinear pooling for person reidentification

Cluster-Based Distribution Alignment For Generalizable Person Re-Identification

Learning to generalize aerial person re‐identification using the meta‐transfer method

Contact Info

Product

Resources

About