Problem
Research questions and friction points this paper is trying to address.
Continual Learning
Online Versatile Incremental Learning
Domain Shift
Class Shift
Dynamic Environments
Online Versatile Incremental Learning: Towards Class and Domain-Agnostic Adaptation at Any Time
🤖 AI Summary
This study addresses the failure of existing methods under the simultaneous online evolution of categories and domains by introducing, for the first time, the Online Visual Incremental Learning (Online VIL) scenario and proposing the TopFlow framework. TopFlow integrates Diffusion Flow Matching (DFM) with Geodesic Topology Preservation (GTP) mechanisms, synergizing contrastive learning and global feature space structure preservation techniques to achieve synchronized class-domain adaptation without explicit boundaries. Experimental results demonstrate that TopFlow attains state-of-the-art performance in dynamic environments, effectively overcoming the limitations inherent in traditional incremental learning paradigms.
📝 Abstract
Continual learning enables vision systems to adapt to ever-changing data distributions. Despite significant advances, existing approaches fail to capture continuous and concurrent shifts in classes and domains, a critical capability for real-world deployment. This work introduces Online VIL (Online Versatile Incremental Learning), a novel scenario where class concepts and visual domains evolve simultaneously online without explicit boundaries. To better adapt to the challenges of such dynamic environments that more closely resemble real-world conditions, we propose a novel framework TopFlow, Topology preservation with Flow matching representation that contains two complementary mechanisms: Domain-agnostic Flow Matching (DFM) and Global Topology Preservation (GTP). DFM guides the model to have domain-agnostic representations by integrating the geodesic flow kernel into contrastive learning. In contrast, GTP maintains the global structure of the feature space without explicitly storing past examples. Our extensive experiments demonstrate that TopFlow effectively addresses the limitations of existing methods within the Online VIL scenario, achieving state-of-the-art performance in challenging Online VIL. The proposed methods suggest potential directions for building continual learning systems in realistic dynamic environments. Our implementation code is available at https://github.com/KU-VGI/Online-VIL.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Online Versatile Incremental Learning
Flow Matching
Topology Preservation
Contrastive Learning
Continual Learning
🔎 Similar Papers
No similar papers found.
