Debiased Batch Normalization via Gaussian Process for Generalizable Person Re-identification

Abstract

Generalizable person re-identification aims to learn a model with only several labeled source domains that can perform well on unseen domains. Without access to the unseen domain, the feature statistics of the batch normalization (BN) layer learned from a limited number of source domains is doubtlessly biased for unseen domain. This would mislead the feature representation learning for unseen domain and deteriorate the generalizaiton ability of the model. In this paper, we propose a novel Debiased Batch Normalization via Gaussian Process approach (GDNorm) for generalizable person re-identification, which models the feature statistic estimation from BN layers as a dynamically self-refining Gaussian process to alleviate the bias to unseen domain for improving the generalization. Specifically, we establish a lightweight model with multiple set of domain-specific BN layers to capture the discriminability of individual source domain, and learn the corresponding parameters of the domain-specific BN layers. These parameters of different source domains are employed to deduce a Gaussian process. We randomly sample several paths from this Gaussian process served as the BN estimations of potential new domains outside of existing source domains, which can further optimize these learned parameters from source domains, and estimate more accurate Gaussian process by them in return, tending to real data distribution. Even without a large number of source domains, GDNorm can still provide debiased BN estimation by using the mean path of the Gaussian process, while maintaining low computational cost during testing. Extensive experiments demonstrate that our GDNorm effectively improves the generalization ability of the model on unseen domain.

Sustainable Isopropyl Alcohol Recovery via Data-Driven, Active-Learning Optimization of Vacuum Membrane Distillation

“…After obtaining initial data points, Bayesian optimization (BO) was employed for active learning. All experimental results were standardized to have zero mean and unit variance, a step crucial for efficient Gaussian Process (GP) modeling . A holistic illustration of this iterative decision-making scheme of BO can be seen in Figure .…”

“…All experimental results were standardized to have zero mean and unit variance, a step crucial for efficient Gaussian Process (GP) modeling. 28 A holistic illustration of this iterative decision-making scheme of BO can be seen in Figure 3. A GP with a Radial Basis Function (RBF) kernel was chosen for surrogate modeling.…”

A method based on difference guide and feature self-enhancement for clothes-changing person re-identification

“…To solve the above problems, [17] propose a novel method called the relevanceaware mixture of experts (RaMoE), using an effective voting-based mixture mechanism to dynamically leverage source domains' diverse characteristics to improve the model's generalization. [18] proposed a novel Debiased Batch Normalization via Gaussian Process approach (GDNorm) for generalizable person ReID, which models the feature statistic estimation from BN layers as a dynamically self-refining Gaussian process to alleviate the bias to unseen domain for improving the generalization. [19] found that the presence of Noisy Annotation (NA) will lead to the effect of noisy annotations on the constructed dataset, To solve the problem, a novel method for robust Visible-Infrared person re-identification, termed Du-Ally Robust Training (DART) was proposed.…”

Domain Generalizable Person Search Using Unreal Dataset

“…Choi et al adopted the batch-instance normalization layers trained with the meta-learning strategy to avoid the overfitting to the source domain (Choi et al 2021). Liu et al defined a hybrid domain composed of the datasets from multiple domains, and trained the dataset in the hybrid domain with an ensemble of other batch normalization parameters to encourage the generalization capability (Liu et al 2022). However, these methods were devised for re-identification and cannot be directly applied to the person search task combined with the addition person detection.…”