FOCUS: Bridging Fine-Grained Recognition and Open-World Discovery across Domains

TL;DR

This paper presents FoCUS, a unified framework for fine-grained domain-generalized category discovery that effectively recognizes known and discovers novel classes across unseen domains, with new benchmarks and improved performance.

Contribution

The paper introduces the first FG-DG-GCD framework, new benchmarks with stylized domains, and combines parts discovery with uncertainty-aware feature augmentation for better open-world recognition.

Findings

FoCUS outperforms existing baselines in clustering accuracy.

It achieves nearly 3x higher computational efficiency.

The benchmarks facilitate systematic evaluation of fine-grained open-world recognition.

Abstract

We introduce the first unified framework for *Fine-Grained Domain-Generalized Generalized Category Discovery* (FG-DG-GCD), bringing open-world recognition closer to real-world deployment under domain shift. Unlike conventional GCD, which assumes labeled and unlabeled data come from the same distribution, DG-GCD learns only from labeled source data and must both recognize known classes and discover novel ones in unseen, unlabeled target domains. This problem is especially challenging in fine-grained settings, where subtle inter-class differences and large intra-class variation make domain generalization significantly harder. To support systematic evaluation, we establish the first *FG-DG-GCD benchmarks* by creating identity-preserving *painting* and *sketch* domains for CUB-200-2011, Stanford Cars, and FGVC-Aircraft using controlled diffusion-adapter stylization. On top of this ,we propose FoCUS, a single-stage framework that combines *Domain-Consistent Parts Discovery* (DCPD) for geometry-stable part reasoning with *Uncertainty-Aware Feature Augmentation* (UFA) for confidence-calibrated feature regularization through uncertainty-guided perturbations. Extensive experiments show that FoCUS outperforms strong GCD, FG-GCD, and DG-GCD baselines by **3.28%**, **9.68%**, and **2.07%**, respectively, in clustering accuracy on the proposed benchmarks. It also remains competitive on coarse-grained DG-GCD tasks while achieving nearly **3x** higher computational efficiency than the current state of the art. ^[Code and datasets will be released upon acceptance.]