Mutual Mean-Teaching: Pseudo Label Refinery for Unsupervised Domain Adaptation on Person Re-identification

By Yixiao Ge, Dapeng Chen, Hongsheng Li. 2020

663 citations recorded by the source corpus.

Person re-identification (re-ID) aims at identifying the same persons' images across different cameras. However, domain diversities between different datasets pose an evident challenge for adapting the re-ID model trained on one dataset to another one. State-of-the-art unsupervised domain adaptation methods for person re-ID transferred the learned knowledge from the source domain by optimizing with pseudo labels created by clustering algorithms on the target domain. Although they achieved state-of-the-art performances, the inevitable label noise caused by the clustering procedure was ignored. Such noisy pseudo labels substantially hinders the model's capability on further improving feature representations on the target domain. In order to mitigate the effects of noisy pseudo labels, we propose to softly refine the pseudo labels in the target domain by proposing an unsupervised framework, Mutual Mean-Teaching (MMT), to learn better features from the target domain via off-line refined hard pseudo labels and on-line refined soft pseudo labels in an alternative training manner. In addition, the common practice is to adopt both the classification loss and the triplet loss jointly for achieving optimal performances in person re-ID models. However, conventional triplet loss cannot work with softly refined labels. To solve this problem, a novel soft softmax-triplet loss is proposed to support learning with soft pseudo triplet labels for achieving the optimal domain adaptation performance. The proposed MMT framework achieves considerable improvements of 14.4%, 18.2%, 13.1% and 16.4% mAP on Market-to-Duke, Duke-to-Market, Market-to-MSMT and Duke-to-MSMT unsupervised domain adaptation tasks.

Subfields: Computer Vision and Pattern Recognition

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

Citations in the corpus

Continual Learning on Noisy Data Streams via Self-Purified Replay (2021, background): “Small loss sample selection [36, 72, 45] techniques by co-teaching [85, 21, 26, 90, 55, 11] identify noisy samples with multiple models in the same vein.”
Layerwise Optimization by Gradient Decomposition for Continual Learning (2021, neutral): “Layerwise gradient update strategy is also applied in meta-learning for rapid loss convergence on transfer learning [13, 14, 48] and few-shot learning [49, 28].”
Patch-based Knowledge Distillation for Lifelong Person Re-Identification (2022, background): “Person ReID has made remarkable progress under various settings such as supervised learning [18, 40, 57], unsupervised learning [9, 36] and unsupervised domain adaptation [12, 15, 66, 75, 76].”
A Survey on Negative Transfer (2020, background): “Target prediction enhancement Soft labeling [135]–[137]”
Discriminability Distillation in Group Representation Learning (2020, background): “It is also beneficial to perform person re-identification [61,39,18,16,17,33,15] from videos.”
Cluster Contrast for Unsupervised Person Re-Identification (2021, methodology): “In contrast to the previous instance-level feature memory methods [11, 12], the P × Z query instance features were just updated in the instance level memory through Eq.”
Debiased Pseudo Labeling in Self-Training (2022, background): “, 2020b), or from another model in an online mutual-teaching manner (Ge et al., 2020).”
Group-aware Label Transfer for Domain Adaptive Person Re-identification (2021, neutral): “Clustering-based Methods in general generate hard or soft pseudo labels based on clustering results and then ﬁne-tune/train the models based on the pseudo labels [9, 36, 31, 10, 33, 40, 14, 2, 6].”