Saved in:
Bibliographic Details
Main Authors: Yu, Hao, Du, Yingxiao, Wu, Jianxin
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2401.16811
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917577991651328
author Yu, Hao
Du, Yingxiao
Wu, Jianxin
author_facet Yu, Hao
Du, Yingxiao
Wu, Jianxin
contents The training datasets used in long-tailed recognition are extremely unbalanced, resulting in significant variation in per-class accuracy across categories. Prior works mostly used average accuracy to evaluate their algorithms, which easily ignores those worst-performing categories. In this paper, we aim to enhance the accuracy of the worst-performing categories and utilize the harmonic mean and geometric mean to assess the model's performance. We revive the balanced undersampling idea to achieve this goal. In few-shot learning, balanced subsets are few-shot and will surely under-fit, hence it is not used in modern long-tailed learning. But, we find that it produces a more equitable distribution of accuracy across categories with much higher harmonic and geometric mean accuracy, and, but lower average accuracy. Moreover, we devise a straightforward model ensemble strategy, which does not result in any additional overhead and achieves improved harmonic and geometric mean while keeping the average accuracy almost intact when compared to state-of-the-art long-tailed learning methods. We validate the effectiveness of our approach on widely utilized benchmark datasets for long-tailed learning. Our code is at \href{https://github.com/yuhao318/BTM/}{https://github.com/yuhao318/BTM/}.
format Preprint
id arxiv_https___arxiv_org_abs_2401_16811
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Reviving Undersampling for Long-Tailed Learning
Yu, Hao
Du, Yingxiao
Wu, Jianxin
Computer Vision and Pattern Recognition
The training datasets used in long-tailed recognition are extremely unbalanced, resulting in significant variation in per-class accuracy across categories. Prior works mostly used average accuracy to evaluate their algorithms, which easily ignores those worst-performing categories. In this paper, we aim to enhance the accuracy of the worst-performing categories and utilize the harmonic mean and geometric mean to assess the model's performance. We revive the balanced undersampling idea to achieve this goal. In few-shot learning, balanced subsets are few-shot and will surely under-fit, hence it is not used in modern long-tailed learning. But, we find that it produces a more equitable distribution of accuracy across categories with much higher harmonic and geometric mean accuracy, and, but lower average accuracy. Moreover, we devise a straightforward model ensemble strategy, which does not result in any additional overhead and achieves improved harmonic and geometric mean while keeping the average accuracy almost intact when compared to state-of-the-art long-tailed learning methods. We validate the effectiveness of our approach on widely utilized benchmark datasets for long-tailed learning. Our code is at \href{https://github.com/yuhao318/BTM/}{https://github.com/yuhao318/BTM/}.
title Reviving Undersampling for Long-Tailed Learning
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2401.16811