COCA: Classifier-Oriented Calibration via Textual Prototype for Source-Free Universal Domain Adaptation
This addresses data efficiency for domain adaptation researchers, though it is incremental as it builds on existing VLM and few-shot learning approaches.
The paper tackles the labeling cost issue in source-free universal domain adaptation by introducing COCA, a plug-and-play method using textual prototypes with vision-language models, which outperforms state-of-the-art models in experiments.
Universal domain adaptation (UniDA) aims to address domain and category shifts across data sources. Recently, due to more stringent data restrictions, researchers have introduced source-free UniDA (SF-UniDA). SF-UniDA methods eliminate the need for direct access to source samples when performing adaptation to the target domain. However, existing SF-UniDA methods still require an extensive quantity of labeled source samples to train a source model, resulting in significant labeling costs. To tackle this issue, we present a novel plug-and-play classifier-oriented calibration (COCA) method. COCA, which exploits textual prototypes, is designed for the source models based on few-shot learning with vision-language models (VLMs). It endows the VLM-powered few-shot learners, which are built for closed-set classification, with the unknown-aware ability to distinguish common and unknown classes in the SF-UniDA scenario. Crucially, COCA is a new paradigm to tackle SF-UniDA challenges based on VLMs, which focuses on classifier instead of image encoder optimization. Experiments show that COCA outperforms state-of-the-art UniDA and SF-UniDA models.