Rethinking Class-Balanced Methods for Long-Tailed Visual Recognition From a Domain Adaptation Perspective

Jamal, Muhammad Abdullah; Brown, Matthew A.; Yang, Ming–Hsuan; Wang, Liqiang; Gong, Boqing

doi:10.1109/cvpr42600.2020.00763

Cited by 206 publications

(176 citation statements)

References 25 publications

Supporting

Mentioning

153

Contrasting

Order By: Relevance

“…Denoting the two domains as D 𝑢 and D 𝑏 , their distributions in feature space are defined as D𝑢 and D𝑏 , respectively. Notably, we not only consider the long-tailed recognition problem as a domain adaptation task, analogous to [14], but also propose to define and slack the generalization error bound.…”

Section: Methodology 31 Modelingmentioning

confidence: 99%

“…The real-world data, however, typically follows longtailed distributions [7,9] To adapt the models trained upon long-tailed distributions to uniform distributions, re-sampling and re-weighting methods [3,6] have been explored to re-balance the data distributions. Recently, feature-classifier decoupling learning strategies [14,16,23,33] are proposed to learn representation and classifiers, step by step. Despite of the effectiveness, these approaches require complicated training strategies and/or work in empirical fashions, which hinder the interpretability and further progress of long-tailed recognition problems.…”

Section: Introductionmentioning

confidence: 99%

“…In this study, analogous to [14], we treat the unbalanced training set as an unbalanced domain and the balanced test set as a balanced domain, solving the Long-tailed recognition problem in a Domain Adaptation (LDA) fashion. Based on the upper bound of generalization error of domain adaptation [1,2], LDA firstly defines the relaxed upper bound of generalization error for the balanced domain.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Long-tailed Distribution Adaptation

Peng

Huang

Guo

et al. 2021

Proceedings of the 29th ACM International Conference on Multimedia

View full text Add to dashboard Cite

Recognizing images with long-tailed distributions remains a challenging problem while there lacks an interpretable mechanism to solve this problem. In this study, we formulate Long-tailed recognition as Domain Adaption (LDA), by modeling the long-tailed distribution as an unbalanced domain and the general distribution as a balanced domain. Within the balanced domain, we propose to slack the generalization error bound, which is defined upon the empirical risks of unbalanced and balanced domains and the divergence between them. We propose to jointly optimize empirical risks of the unbalanced and balanced domains and approximate their domain divergence by intra-class and inter-class distances, with the aim to adapt models trained on the long-tailed distribution to general distributions in an interpretable way. Experiments on benchmark datasets for image recognition, object detection, and instance segmentation validate that our LDA approach, beyond its interpretability, achieves state-of-the-art performance.

show abstract

Section: Methodology 31 Modelingmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Long-tailed Distribution Adaptation

Peng

Huang

Guo

et al. 2021

Proceedings of the 29th ACM International Conference on Multimedia

View full text Add to dashboard Cite

show abstract

“…In dealing with the problem of dataset imbalance, we can also learn from other learning strategies. With meta learning (domain adaptation), minor categories and major categories are processed differently to learn how to reweight adaptively (Shu et al, 2019 ), or to formulate domain adaptation problems (Jamal et al, 2020 ). Metric learning essentially models the boundary/margin near minor categories, with the aim of learning better embedding (Huang et al, 2016 ; Zhang et al, 2017 ).…”

Section: Related Workmentioning

confidence: 99%

Convolutional Rebalancing Network for the Classification of Large Imbalanced Rice Pest and Disease Datasets in the Field

et al. 2021

View full text Add to dashboard Cite

The accurate classification of crop pests and diseases is essential for their prevention and control. However, datasets of pest and disease images collected in the field usually exhibit long-tailed distributions with heavy category imbalance, posing great challenges for a deep recognition and classification model. This paper proposes a novel convolutional rebalancing network to classify rice pests and diseases from image datasets collected in the field. To improve the classification performance, the proposed network includes a convolutional rebalancing module, an image augmentation module, and a feature fusion module. In the convolutional rebalancing module, instance-balanced sampling is used to extract features of the images in the rice pest and disease dataset, while reversed sampling is used to improve feature extraction of the categories with fewer images in the dataset. Building on the convolutional rebalancing module, we design an image augmentation module to augment the training data effectively. To further enhance the classification performance, a feature fusion module fuses the image features learned by the convolutional rebalancing module and ensures that the feature extraction of the imbalanced dataset is more comprehensive. Extensive experiments in the large-scale imbalanced dataset of rice pests and diseases (18,391 images), publicly available plant image datasets (Flavia, Swedish Leaf, and UCI Leaf) and pest image datasets (SMALL and IP102) verify the robustness of the proposed network, and the results demonstrate its superior performance over state-of-the-art methods, with an accuracy of 97.58% on rice pest and disease image dataset. We conclude that the proposed network can provide an important tool for the intelligent control of rice pests and diseases in the field.

show abstract

“…Instead of hacking the prior distribution of classes, focusing on the hard samples also alleviates the long-tailed issue. Jamal et al [174] adopt meta-learning to dynamically learn the weights for each instance by iteratively optimize an inner loop. Li et al [175] enhance the hard samples of by group softmax.…”

Section: Hard Example Miningmentioning

confidence: 99%

Towards robust inference against distribution shifts in computer vision

Tang¹

View full text Add to dashboard Cite

Hanwang Zhang. Without his patient guidance and continuous help, I would never have accomplished these research projects and published corresponding papers. He opened the door of academic research for me with his invaluable expertise. Before I met him, I was just a green hand that had a pretty bad taste of research and zero experience of publishing top-tier papers. However, he reinvented me with his immense knowledge and insightful support, which guided me to find my research topics, sharpen my critical thinking, build my academic skills and bring my work to a higher level. I would like to acknowledge my collaborators, Dr.

show abstract

Rethinking Class-Balanced Methods for Long-Tailed Visual Recognition From a Domain Adaptation Perspective

Cited by 206 publications

References 25 publications

Long-tailed Distribution Adaptation

Long-tailed Distribution Adaptation

Convolutional Rebalancing Network for the Classification of Large Imbalanced Rice Pest and Disease Datasets in the Field

Towards robust inference against distribution shifts in computer vision

Contact Info

Product

Resources

About