Class Instance Balanced Learning for Long-Tailed Classification

TL;DR

This paper introduces a class instance balanced loss (CIBL) that dynamically reweights cross-entropy and contrastive losses based on class frequency, improving long-tailed classification performance and allowing performance tradeoffs between common and rare classes.

Contribution

The paper proposes a novel CIBL method that balances class contributions in long-tailed learning, and demonstrates its effectiveness on benchmark datasets with faster training using a cosine classifier.

Findings

CIBL improves class balance in long-tailed datasets.

Adjusting contrastive weight shifts performance between head and tail classes.

Cosine classifier reduces training epochs while maintaining accuracy.

Abstract

The long-tailed image classification task remains important in the development of deep neural networks as it explicitly deals with large imbalances in the class frequencies of the training data. While uncommon in engineered datasets, this imbalance is almost always present in real-world data. Previous approaches have shown that combining cross-entropy and contrastive learning can improve performance on the long-tailed task, but they do not explore the tradeoff between head and tail classes. We propose a novel class instance balanced loss (CIBL), which reweights the relative contributions of a cross-entropy and a contrastive loss as a function of the frequency of class instances in the training batch. This balancing favours the contrastive loss for more common classes, leading to a learned classifier with a more balanced performance across all class frequencies. Furthermore, increasing the relative weight on the contrastive head shifts performance from common (head) to rare (tail) classes, allowing the user to skew the performance towards these classes if desired. We also show that changing the linear classifier head with a cosine classifier yields a network that can be trained to similar performance in substantially fewer epochs. We obtain competitive results on both CIFAR-100-LT and ImageNet-LT.