Curvature-Balanced Feature Manifold Learning for Long-Tailed Classification

arXiv:2303.12307v3 Announce Type: replace
Abstract: To address the challenges of long-tailed classification, researchers have proposed several approaches to reduce model bias, most of which assume that classes with few samples are weak classes. However, recent studies have shown that tail classes are not always hard to learn, and model bias has been observed on sample-balanced datasets, suggesting the existence of other factors that affect model bias. In this work, we systematically propose a series of geometric measurements for perceptual manifolds …

abstract arxiv bias challenges classification cs.ai cs.cv feature however learn manifold model bias reduce researchers samples studies type