Learning to Generalize Unseen Domains via Memory-based Multi-Source Meta-Learning for Person Re-Identification

@article{Zhao2021LearningTG,
  title={Learning to Generalize Unseen Domains via Memory-based Multi-Source Meta-Learning for Person Re-Identification},
  author={Yuyang Zhao and Zhun Zhong and Fengxiang Yang and Zhiming Luo and Yaojin Lin and Shaozi Li and N. Sebe},
  journal={2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  year={2021},
  pages={6273-6282}
}
  • Yuyang Zhao, Zhun Zhong, N. Sebe
  • Published 1 December 2020
  • Computer Science
  • 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
Recent advances in person re-identification (ReID) obtain impressive accuracy in the supervised and unsupervised learning settings. However, most of the existing methods need to train a new model for a new domain by accessing data. Due to public privacy, the new domain data are not always accessible, leading to a limited applicability of these methods. In this paper, we study the problem of multi-source domain generalization in ReID, which aims to learn a model that can perform well on unseen… 

Figures and Tables from this paper

Unsupervised Domain Generalization for Person Re-identification: A Domain-specific Adaptive Framework
TLDR
This work proposes a simple and efficient domain-specific adaptive framework, and realizes it with an adaptive normalization module designed upon the batch and instance normalization techniques, which successfully yield reliable pseudo-labels to implement training and also enhance the domain generalization capability of the model as required.
Adaptive Domain-Specific Normalization for Generalizable Person Re-Identification
TLDR
This work proposes a novel adaptive domain-specific normalization approach (AdsNorm) for generalizable person Re-ID, which describes unseen target domain as a combination of the known source ones, and explicitly learns domain- specific representation with target distribution to improve the model’s generalization by a meta-learning pipeline.
Multiple Domain Experts Collaborative Learning: Multi-Source Domain Generalization For Person Re-Identification
TLDR
Extensive experiments on DG-ReID benchmarks show that the MD-ExCo outperforms the state-of-the-art methods by a large margin, showing its ability to improve the generalization capability of the ReID models.
TAL: Two-stream Adaptive Learning for Generalizable Person Re-identification
TLDR
This work argues that both domain-specific and domain-invariant features are crucial for improving the generalization ability of re-id models, and designs a novel framework, which is name two-stream adaptive learning (TAL), to simultaneously model these two kinds of information.
META: Mimicking Embedding via oThers' Aggregation for Generalizable Person Re-identification
TLDR
This paper presents a new approach called Mimicking Embedding via oThers’ Aggregation (META), which surpasses state-of-the-art DG ReID methods by a large margin and can expect META to mimic embedding for a truly unseen target domain.
Semi-Supervised Domain Generalizable Person Re-Identification
TLDR
This work aims to explore multiple labeled datasets to learn generalized domain-invariant representations for person re-id, and proposes a simple yet effective Semi-Supervised Knowledge Distillation (SSKD) framework that effectively exploits the weakly annotated data by assigning soft pseudo labels to YouTube-Human for improving the generalization ability of models.
Unleashing the Potential of Unsupervised Pre-Training with Intra-Identity Regularization for Person Re-Identification
TLDR
An Unsupervised Pre-training framework for ReID based on the contrastive learning (CL) pipeline, dubbed UP-ReID, which is instantiated as two constraints coming from global image aspect and local patch aspect, which can significantly benefit the downstream ReID fine-tuning and achieve state-of-the-art performance.
DEX: Domain Embedding Expansion for Generalized Person Re-identification
TLDR
This paper focuses on developing a domain generalization (DG) Person ReID model that can be deployed without additional fine-tuning or adaptation, and proposes the Domain Embedding Expansion (DEX) module, which dynamically manipulates and augments deep features based on person and domain labels during training, significantly improving the generalization capability and robustness of Person Re ID models to unseen domains.
Domain Generalization in Vision: A Survey
TLDR
A comprehensive literature review is provided to summarize the developments in DG for computer vision over the past decade and conducts a thorough review into existing methods and presents a categorization based on their methodologies and motivations.
Calibrated Feature Decomposition for Generalizable Person Re-Identification
TLDR
A simple yet effective Calibrated Feature Decomposition (CFD) module that focuses on improving the generalization capacity for person re-identification through a more judicious feature decomposition and reinforcement strategy.
...
1
2
3
4
...

References

SHOWING 1-10 OF 60 REFERENCES
Generalizable Person Re-Identification by Domain-Invariant Mapping Network
TLDR
DIMN is designed to learn a mapping between a person image and its identity classifier, i.e., it produces a classifier using a single shot and can be used to match an arbitrary number of identities in a target domain.
Learning to Adapt Invariance in Memory for Person Re-Identification
TLDR
This work considers the problem of unsupervised domain adaptation in person re-identification (re-ID), which aims to transfer knowledge from the source domain to the target domain, and proposes a novel adaptation framework w.r.t three types of underlying invariance, i.e., Exemplar-Invariance, Camera-Invarance, and Neighborhood- Invariance.
Learning Meta Face Recognition in Unseen Domains
TLDR
This paper proposes a novel face recognition method via meta-learning named Meta Face Recognition (MFR), which synthesizes the source/target domain shift with a meta-optimization objective, which requires the model to learn effective representations not only on synthesized source domains but also on synthesizer target domains.
Generalizing a Person Retrieval Model Hetero- and Homogeneously
TLDR
A Hetero-Homogeneous Learning (HHL) method is introduced that enforces two properties simultaneously: camera invariance, learned via positive pairs formed by unlabeled target images and their camera style transferred counterparts and domain connectedness, implemented by homogeneous learning.
Joint Disentangling and Adaptation for Cross-Domain Person Re-Identification
TLDR
This paper proposes a joint learning framework that disentangles id-related/unrelated features and enforces adaptation to work on the id- related feature space exclusively, and outperforms the state-of-the-art methods by clear margins.
Self-Similarity Grouping: A Simple Unsupervised Cross Domain Adaptation Approach for Person Re-Identification
TLDR
A Self-similarity Grouping (SSG) approach, which exploits the potential similarity of unlabeled samples to build multiple clusters from different views automatically, and introduces a clustering-guided semisupervised approach named SSG ++ to conduct the one-shot domain adaption in an open set setting.
AD-Cluster: Augmented Discriminative Clustering for Domain Adaptive Person Re-Identification
TLDR
A novel augmented discriminative clustering (AD-Cluster) technique that estimates and augments person clusters in target domains and enforces the discrimination ability of re-ID models with the augmented clusters.
Invariance Matters: Exemplar Memory for Domain Adaptive Person Re-Identification
TLDR
Results on three re-ID domains show that the domain adaptation accuracy outperforms the state of the art by a large margin and the three invariance properties and the proposed memory are indispensable towards an effective domain adaptation system.
Unsupervised Person Re-identification: Clustering and Fine-tuning
TLDR
A progressive unsupervised learning (PUL) method to transfer pretrained deep representations to unseen domains and demonstrates that PUL outputs discriminative features that improve the re-ID accuracy.
Style Normalization and Restitution for Generalizable Person Re-Identification
TLDR
The aim of this paper is to design a generalizable person ReID framework which trains a model on source domains yet is able to generalize/perform well on target domains, and to enforce a dual causal loss constraint in SNR to encourage the separation of identity-relevant features and identity-irrelevant features.
...
1
2
3
4
5
...