Adjusting Logit in Gaussian Form for Long-Tailed Visual Recognition

TL;DR

This paper proposes a Gaussian-based feature augmentation and logit adjustment method to balance embedding spaces in long-tailed visual recognition, improving classifier performance with minimal computational cost.

Contribution

It introduces a novel feature augmentation technique and logit adjustment methods to calibrate embedding spaces in long-tailed data, enhancing classifier accuracy.

Findings

Outperforms state-of-the-art methods on benchmark datasets.

Balances embedding space distribution effectively.

Reduces classifier bias with modest computational overhead.

Abstract

It is not uncommon that real-world data are distributed with a long tail. For such data, the learning of deep neural networks becomes challenging because it is hard to classify tail classes correctly. In the literature, several existing methods have addressed this problem by reducing classifier bias, provided that the features obtained with long-tailed data are representative enough. However, we find that training directly on long-tailed data leads to uneven embedding space. That is, the embedding space of head classes severely compresses that of tail classes, which is not conducive to subsequent classifier learning. This paper therefore studies the problem of long-tailed visual recognition from the perspective of feature level. We introduce feature augmentation to balance the embedding distribution. The features of different classes are perturbed with varying amplitudes in Gaussian form. Based on these perturbed features, two novel logit adjustment methods are proposed to improve model performance at a modest computational overhead. Subsequently, the distorted embedding spaces of all classes can be calibrated. In such balanced-distributed embedding spaces, the biased classifier can be eliminated by simply retraining the classifier with class-balanced sampling data. Extensive experiments conducted on benchmark datasets demonstrate the superior performance of the proposed method over the state-of-the-art ones. Source code is available at https://github.com/Keke921/GCLLoss.