DualCP: Rehearsal-Free Domain-Incremental Learning via Dual-Level Concept Prototype
This addresses a practical challenge in deploying AI systems in real-world environments with changing conditions while respecting privacy constraints, though it is an incremental improvement over existing methods.
The paper tackles the problem of rehearsal-free domain-incremental learning for vision models, where models must adapt to new domains without storing old data, and achieves state-of-the-art performance on benchmarks like DomainNet, CDDB, and CORe50.
Domain-Incremental Learning (DIL) enables vision models to adapt to changing conditions in real-world environments while maintaining the knowledge acquired from previous domains. Given privacy concerns and training time, Rehearsal-Free DIL (RFDIL) is more practical. Inspired by the incremental cognitive process of the human brain, we design Dual-level Concept Prototypes (DualCP) for each class to address the conflict between learning new knowledge and retaining old knowledge in RFDIL. To construct DualCP, we propose a Concept Prototype Generator (CPG) that generates both coarse-grained and fine-grained prototypes for each class. Additionally, we introduce a Coarse-to-Fine calibrator (C2F) to align image features with DualCP. Finally, we propose a Dual Dot-Regression (DDR) loss function to optimize our C2F module. Extensive experiments on the DomainNet, CDDB, and CORe50 datasets demonstrate the effectiveness of our method.