Exploring Sparse Visual Prompt for Domain Adaptive Dense Prediction

1Peking University, 2Harbin Institute of Technology, Shenzhen 3Beihang University 4University of Science and Technology of China
* indicates Equal Contribution


The visual prompts have provided an efficient manner in addressing visual cross-domain problems. In previous works, (Gan et al. 2022a) first introduces domain prompts to tackle the classification Test-Time Adaptation (TTA) problem by warping image-level prompts on the input and fine-tuning prompts for each target domain. However, since the imagelevel prompts mask out continuous spatial details in the prompt-allocated region, it will suffer from inaccurate contextual information and limited domain knowledge extraction, particularly when dealing with dense prediction TTA problems. To overcome these challenges, we propose a novel Sparse Visual Domain Prompts (SVDP) approach, which holds minimal trainable parameters (e.g., 0.1%) in the imagelevel prompt and reserves more spatial information of the input. To better apply SVDP in extracting domain-specific knowledge, we introduce the Domain Prompt Placement (DPP) method to adaptively allocates trainable parameters of SVDP on the pixels with large distribution shifts. Furthermore, recognizing that each target domain sample exhibits a unique domain shift, we design Domain Prompt Updating (DPU) strategy to optimize prompt parameters differently for each sample, facilitating efficient adaptation to the target domain. Extensive experiments were conducted on widelyused TTA and continual TTA benchmarks, and our proposed method achieves state-of-the-art performance in both semantic segmentation and depth estimation tasks


Related Links

There's a lot of excellent work that was published related to the Test-Time Adaptation and Continual Test-Time Adaptation,.


      title={Exploring Sparse Visual Prompt for Domain Adaptive Dense Prediction},
      author={Yang, Senqiao and Wu, Jiarui and Liu, Jiaming and Li, Xiaoqi and Zhang, Qizhe and Pan, Mingjie and Gan, Yulu and Chen, Zehui and Zhang, Shanghang},
      booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},