Fine-grained Classes and How to Find Them

TL;DR

FALCON is an unsupervised method that discovers fine-grained classes from coarse labels, effectively inferring subtle class differences without fine-grained annotations, and works across multiple datasets.

Contribution

The paper introduces FALCON, a novel modular approach that infers fine-grained classes from coarse labels without supervision, handling multiple datasets with different labeling strategies.

Findings

FALCON outperforms baselines by a large margin.

Achieves 22% improvement on tieredImageNet dataset.

Successfully applied to image and single-cell classification tasks.

Abstract

In many practical applications, coarse-grained labels are readily available compared to fine-grained labels that reflect subtle differences between classes. However, existing methods cannot leverage coarse labels to infer fine-grained labels in an unsupervised manner. To bridge this gap, we propose FALCON, a method that discovers fine-grained classes from coarsely labeled data without any supervision at the fine-grained level. FALCON simultaneously infers unknown fine-grained classes and underlying relationships between coarse and fine-grained classes. Moreover, FALCON is a modular method that can effectively learn from multiple datasets labeled with different strategies. We evaluate FALCON on eight image classification tasks and a single-cell classification task. FALCON outperforms baselines by a large margin, achieving 22% improvement over the best baseline on the tieredImageNet dataset with over 600 fine-grained classes.