Visual Prompt Tuning For Test-time Domain Adaptation

Gao Yunhe, Shi Xingjian, Zhu Yi, Wang Hao, Tang Zhiqiang, Zhou Xiong, Li Mu, Metaxas Dimitris N.. Arxiv 2022

[Paper]    
Efficiency And Optimization Fine Tuning Model Architecture Multimodal Models Pretraining Methods Prompting Reinforcement Learning Training Techniques Transformer

Models should be able to adapt to unseen data during test-time to avoid performance drops caused by inevitable distribution shifts in real-world deployment scenarios. In this work, we tackle the practical yet challenging test-time adaptation (TTA) problem, where a model adapts to the target domain without accessing the source data. We propose a simple recipe called \textit{Data-efficient Prompt Tuning} (DePT) with two key ingredients. First, DePT plugs visual prompts into the vision Transformer and only tunes these source-initialized prompts during adaptation. We find such parameter-efficient finetuning can efficiently adapt the model representation to the target domain without overfitting to the noise in the learning objective. Second, DePT bootstraps the source representation to the target domain by memory bank-based online pseudo-labeling. A hierarchical self-supervised regularization specially designed for prompts is jointly optimized to alleviate error accumulation during self-training. With much fewer tunable parameters, DePT demonstrates not only state-of-the-art performance on major adaptation benchmarks VisDA-C, ImageNet-C, and DomainNet-126, but also superior data efficiency, i.e., adaptation with only 1% or 10% data without much performance degradation compared to 100% data. In addition, DePT is also versatile to be extended to online or multi-source TTA settings.

Similar Work