Feature Generation for Long-tail Classification

TL;DR

This paper proposes a novel feature generation method for long-tail classification that estimates tail class distributions to generate meaningful features, improving classifier training and achieving state-of-the-art results.

Contribution

It introduces a distribution-based feature generation approach for tail classes, decoupling representation and classifier training to enhance long-tail classification performance.

Findings

Achieved new state-of-the-art on CIFAR-100-LT and mini-ImageNet-LT datasets.

Generated features improve classifier training and class representation.

Qualitative analysis confirms the meaningfulness of generated features.

Abstract

The visual world naturally exhibits an imbalance in the number of object or scene instances resulting in a \emph{long-tailed distribution}. This imbalance poses significant challenges for classification models based on deep learning. Oversampling instances of the tail classes attempts to solve this imbalance. However, the limited visual diversity results in a network with poor representation ability. A simple counter to this is decoupling the representation and classifier networks and using oversampling only to train the classifier. In this paper, instead of repeatedly re-sampling the same image (and thereby features), we explore a direction that attempts to generate meaningful features by estimating the tail category's distribution. Inspired by ideas from recent work on few-shot learning, we create calibrated distributions to sample additional features that are subsequently used to train the classifier. Through several experiments on the CIFAR-100-LT (long-tail) dataset with varying imbalance factors and on mini-ImageNet-LT (long-tail), we show the efficacy of our approach and establish a new state-of-the-art. We also present a qualitative analysis of generated features using t-SNE visualizations and analyze the nearest neighbors used to calibrate the tail class distributions. Our code is available at https://github.com/rahulvigneswaran/TailCalibX.