TASIL Pretrained Models
说明
本项目代码已开源并托管于 GitHub：
https://github.com/roomtoor/TASIL主实验对应的预训练模型权重已公开上传至 HuggingFace：
https://huggingface.co/javaccd/TASIL
This repository provides pretrained model checkpoints for TASIL under the Single-Source Domain Generalization (SSDG) setting.
Each checkpoint is trained using one source domain. Together, these checkpoints cover all source-domain choices used in the main experiments.
Office-Home (SSDG)
TACDv2_OfficeHome_3_SSDG_Art_ep25.pth
Trained with Art as the source domain.TACDv2_OfficeHome_3_SSDG_Clipart_ep7.pth
Trained with Clipart as the source domain.TACDv2_OfficeHome_3_SSDG_Product_ep4.pth
Trained with Product as the source domain.TACDv2_OfficeHome_3_SSDG_Real_World_ep9.pth
Trained with Real World as the source domain.
TerraIncognita (DomainBed setting, SSDG)
TACDv2_SSDG_SSDG_terraincognita_location_38_ep17.pth
Trained with location_38 as the source domain.TACDv2_SSDG_SSDG_terraincognita_location_43_ep4.pth
Trained with location_43 as the source domain.TACDv2_SSDG_SSDG_terraincognita_location_46_ep10.pth
Trained with location_46 as the source domain.TACDv2_SSDG_SSDG_terraincognita_location_100_ep1.pth
Trained with location_100 as the source domain.
Notes
- These checkpoints correspond to the main SSDG experiments.
- Training and evaluation scripts are provided in the main project repository.
- Results can be reproduced using the provided commands with fixed random seeds.