The Majority Can Help the Minority: Context-rich Minority Oversampling for Long-tailed Classification

TLDR

Class imbalance degrades classifier generalization because minority classes lack sufficient training data. This work introduces a minority oversampling method that augments minority samples by using majority‑class images as contextual backgrounds. The technique generates diverse synthetic samples by pasting minority‑class images onto majority‑class background images, and it can be combined with existing long‑tailed recognition methods. The method attains state‑of‑the‑art performance on several long‑tailed classification benchmarks without modifying the model architecture. The code is available at https://github.com/naver-ai/cmo.

Abstract

The problem of class imbalanced data is that the gener-alization performance of the classifier deteriorates due to the lack of data from minority classes. In this paper, we pro-pose a novel minority over-sampling method to augment di-versified minority samples by leveraging the rich context of the majority classes as background images. To diversify the minority samples, our key idea is to paste an image from a minority class onto rich-context images from a majority class, using them as background images. Our method is simple and can be easily combined with the existing long-tailed recognition methods. We empirically prove the effectiveness of the proposed oversampling method through extensive experiments and ablation studies. Without any architectural changes or complex algorithms, our method achieves state-of-the-art performance on various long-tailed classification benchmarks. Our code is made available at https://github.com/naver-ai/cmo.