Skip to content

Latest commit

 

History

History
66 lines (44 loc) · 2.85 KB

README.md

File metadata and controls

66 lines (44 loc) · 2.85 KB

HCVP: Leveraging Hierarchical Contrastive Visual Prompt for Domain Generalization

This repository contains the code for the paper "HCVP: Leveraging Hierarchical Contrastive Visual Prompt for Domain Generalization", which is accepted by IEEE Transactions on Multimedia.

About HCVP

The Hierarchical Contrastive Visual Prompt (HCVP) method stands at the forefront of Domain Generalization (DG), leveraging the power of visual prompts to enhance model generalization across unseen domains. HCVP diverges from traditional approaches by introducing domain-level and task-specific prompts, enabling the model to better differentiate between domain-invariant and domain-specific features. For more details, you can find them in our paper.

Overview

🐙 Getting Started

Clone this repository and navigate into the project directory

git clone git@github.com:jameszhou-gl/HCVP
cd HCVP

Create a conda environment and install the required packages

conda create --name hcvp python=3.9.12 -y
conda activate hcvp
pip install -r requirements.txt

🚀 Datasets and model

For detailed instructions, please follow the documents in Domainbed.

python -m domainbed.scripts.download \
       --data_dir=/your-data-path/

Our code is based on DomainBed, you can train and evaluate the model by following the instructions in DomainBed and specify the model class as HCVP.

  • HCVP model class is located in domainbed/algorithms.py including Prompt Contrastive Learning (PCL) and Class-conditional Contrastive Invariance (CCI).
  • Vit backbone is Vit_HCVP in domainbed/networks.py
  • Hierarchical Prompt Generation Network (HPGN) is HierarchicalPromptNetwork in domainbed/prompt_network.py
  • Prompt modulation network (PMN) is SimpleModulationNetwork in domainbed/modulation_network.py

✅ Citation

If you find our work useful in your research, please consider citing:

@misc{zhou2024hcvpleveraginghierarchicalcontrastive,
      title={HCVP: Leveraging Hierarchical Contrastive Visual Prompt for Domain Generalization}, 
      author={Guanglin Zhou and Zhongyi Han and Shiming Chen and Biwei Huang and Liming Zhu and Tongliang Liu and Lina Yao and Kun Zhang},
      year={2024},
      eprint={2401.09716},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2401.09716}, 
}

Acknowlegdement

This code is built on DomainBed, VPT and ood_bench. We thank the authors for sharing their codes.