Invariance Matters: Exemplar Memory for Domain Adaptive Person Re-Identification

By Zhun Zhong, Liang Zheng, Zhiming Luo, Shaozi Li, Yi Yang. Computer Vision and Pattern Recognition, 2019

615 citations recorded by the source corpus.

This paper considers the domain adaptive person re-identification (re-ID) problem: learning a re-ID model from a labeled source domain and an unlabeled target domain. Conventional methods are mainly to reduce feature distribution gap between the source and target domains. However, these studies largely neglect the intra-domain variations in the target domain, which contain critical factors influencing the testing performance on the target domain. In this work, we comprehensively investigate into the intra-domain variations of the target domain and propose to generalize the re-ID model w.r.t three types of the underlying invariance, i.e., exemplar-invariance, camera-invariance and neighborhood-invariance. To achieve this goal, an exemplar memory is introduced to store features of the target domain and accommodate the three invariance properties. The memory allows us to enforce the invariance constraints over global training batch without significantly increasing computation cost. Experiment demonstrates that the three invariance properties and the proposed memory are indispensable towards an effective domain adaptation system. Results on three re-ID domains show that our domain adaptation accuracy outperforms the state of the art by a large margin. Code is available at: https://github.com/zhunzhong07/ECN

Subfields: Computer Vision and Pattern Recognition, Mechanical Engineering

View Paper: https://www.semanticscholar.org/paper/0c5d9f4bf8b11a92a4d42b796052aca20a2a43fd

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Mechanical Engineering

Citations in the corpus

Self-Similarity Grouping: A Simple Unsupervised Cross Domain Adaptation Approach for Person Re-Identification (2018, neutral): “9 ENC [50] 43.”
FSDR: Frequency Space Domain Randomization for Domain Generalization (2021, background): “The third is self-training based which employs ”pseudo label” to guide unsupervised learning over unlabelled target-domain data [88, 57, 83, 19, 86, 17, 24].”
Joint Discriminative and Generative Learning for Person Re-Identification (2019, methodology): “CycleGAN [61] is applied in [9, 60] to transfer pedestrian image style from one dataset to another.”
Deep Learning for Person Re-Identification: A Survey and Outlook (2020, methodology): “It has served as a strong part feature learning baseline in the current state-of-the-art [104], [105], [106].”
Deep Domain-Adversarial Image Generation for Domain Generalisation (2020, background): “It is widely acknowledged that crossdomain re-ID is a challenging problem (Zhong et al. 2019; Zhou et al. 2019a).”
Cluster Contrast for Unsupervised Person Re-Identification (2021, methodology): “The other is the unsupervised domain adaptation (UDA) person re-ID, which first pre-trains a model on the source labeled dataset, and then fine-tunes the model on the target unlabeled dataset [7, 23, 42, 44, 49, 55, 56].”
Infrared-Visible Cross-Modal Person Re-Identification with an X Modality (2020, methodology): “For visible image based deep models, the appearance and colour information usually dominate the learned high-level semantic information (Zhong et al. 2019; Sun and Zheng 2019).”
Hierarchical Clustering With Hard-Batch Triplet Loss for Person Re-Identification (2019, background): “0 ECN[37] Transfer 75.”
Style Normalization and Restitution for Generalizable Person Re-Identification (2020, methodology): “3 ECN [59] CVPR’19 Market1501 + Duke (U) 40.”
Omni-Scale Feature Learning for Person Re-Identification (2019, methodology): “Settings Following the recent works [82, 83, 30, 66, 64], we choose Market1501 and Duke as the target datasets.”
Each Part Matters: Local Patterns Facilitate Cross-View Geo-Localization (2020, methodology): “Currently, some state-of-the-art works [42], [43], [44], [45] extend the PCB with more partitions or optimization losses.”
Group-aware Label Transfer for Domain Adaptive Person Re-identification (2021, methodology): “ECN [55] leverages memory bank to enforce exemplar-invariance, camera-invariance, and neighborhood-invariance over global training data (instead of the local batch) for UDA person ReID.”
Mutual Mean-Teaching: Pseudo Label Refinery for Unsupervised Domain Adaptation on Person Re-identification (2020, background): “ENC (Zhong et al., 2019) assigned soft labels by saving averaged features with an exemplar memory module.”
Taking a Closer Look at Domain Shift: Category-Level Adversaries for Semantics Consistent Domain Adaptation (2018, background): “To address this issue, domain adaptation approaches [34, 40, 14, 45, 17, 16, 13, 47] are proposed to bridge the gap between the source and target domains.”
Unsupervised Person Re-Identification via Multi-Label Classification (2020, methodology): “ECN [46] utilizes transfer learning and minimizes the target invariance.”