Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Generalized Categories Discovery for Long-tailed Recognition

Ziyun Li, Christoph Meinel, Haojin Yang

TL;DR

This work addresses long-tailed Generalized Category Discovery by introducing a lightweight framework that counteracts head-class bias in unlabeled data containing known and unknown categories. It combines tail reweighting and a class prior constraint with moving-average prior estimation and contrastive representation learning, while avoiding iterative optimal-transport steps. The approach minimizes the overall loss $L_{overall}$, comprising $L_{ins}$, $L_{sup}$, and entropy terms $H(q,r)$ and $H(q,u)$, and yields approximately 6–9% improvements on ImageNet100 with competitive results on CIFAR100 under imbalanced conditions. Overall, the method offers a scalable, computation-friendly solution for open-world recognition in long-tailed data regimes.

Abstract

Generalized Class Discovery (GCD) plays a pivotal role in discerning both known and unknown categories from unlabeled datasets by harnessing the insights derived from a labeled set comprising recognized classes. A significant limitation in prevailing GCD methods is their presumption of an equitably distributed category occurrence in unlabeled data. Contrary to this assumption, visual classes in natural environments typically exhibit a long-tailed distribution, with known or prevalent categories surfacing more frequently than their rarer counterparts. Our research endeavors to bridge this disconnect by focusing on the long-tailed Generalized Category Discovery (Long-tailed GCD) paradigm, which echoes the innate imbalances of real-world unlabeled datasets. In response to the unique challenges posed by Long-tailed GCD, we present a robust methodology anchored in two strategic regularizations: (i) a reweighting mechanism that bolsters the prominence of less-represented, tail-end categories, and (ii) a class prior constraint that aligns with the anticipated class distribution. Comprehensive experiments reveal that our proposed method surpasses previous state-of-the-art GCD methods by achieving an improvement of approximately 6 - 9% on ImageNet100 and competitive performance on CIFAR100.

Generalized Categories Discovery for Long-tailed Recognition

TL;DR

This work addresses long-tailed Generalized Category Discovery by introducing a lightweight framework that counteracts head-class bias in unlabeled data containing known and unknown categories. It combines tail reweighting and a class prior constraint with moving-average prior estimation and contrastive representation learning, while avoiding iterative optimal-transport steps. The approach minimizes the overall loss , comprising , , and entropy terms and , and yields approximately 6–9% improvements on ImageNet100 with competitive results on CIFAR100 under imbalanced conditions. Overall, the method offers a scalable, computation-friendly solution for open-world recognition in long-tailed data regimes.

Abstract

Generalized Class Discovery (GCD) plays a pivotal role in discerning both known and unknown categories from unlabeled datasets by harnessing the insights derived from a labeled set comprising recognized classes. A significant limitation in prevailing GCD methods is their presumption of an equitably distributed category occurrence in unlabeled data. Contrary to this assumption, visual classes in natural environments typically exhibit a long-tailed distribution, with known or prevalent categories surfacing more frequently than their rarer counterparts. Our research endeavors to bridge this disconnect by focusing on the long-tailed Generalized Category Discovery (Long-tailed GCD) paradigm, which echoes the innate imbalances of real-world unlabeled datasets. In response to the unique challenges posed by Long-tailed GCD, we present a robust methodology anchored in two strategic regularizations: (i) a reweighting mechanism that bolsters the prominence of less-represented, tail-end categories, and (ii) a class prior constraint that aligns with the anticipated class distribution. Comprehensive experiments reveal that our proposed method surpasses previous state-of-the-art GCD methods by achieving an improvement of approximately 6 - 9% on ImageNet100 and competitive performance on CIFAR100.
Paper Structure (20 sections, 6 equations, 3 figures, 2 tables)

This paper contains 20 sections, 6 equations, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Problem Setup
  3. Method
  4. Class Prior Estimation
  5. Moving-average distribution update
  6. Representation Improvement
  7. Constraints on Class Distribution
  8. Overall Loss Objective
  9. Experiments
  10. Setup
  11. Datasets
  12. Evaluation metrics
  13. Implementation details
  14. Compared with SOTA
  15. Baselines
  16. ...and 5 more sections

Figures (3)

  • Figure 1: Performance trends of Overall, Known, Unknown-Aware, and Unknown-Agnostic accuracies across different $\beta$ values with $\alpha=0$ constant on CIFAR100 with $\rho=5$.
  • Figure 2: Performance trends of Overall, Known, Unknown-Aware, and Unknown-Agnostic accuracies across different $\alpha$ values with $\beta=2$ constant on CIFAR100 with $\rho=5$.
  • Figure 3: Performance trends of Overall, Known, Unknown-Aware, and Unknown-Agnostic accuracies across different $\alpha$ values with $\beta=5$ constant on CIFAR100 with $\rho=5$.

Theorems & Definitions (1)

  • Definition 1: Long-tailed Generalized Category Discovery