Official PyTorch implementation of "Unknown Domain Inconsistency Minimization for Domain Generalization" (ICLR 2024) by Seungjae Shin, HeeSun Bae, Byeonghu Na, Yoon-Yeong Kim, and Il-chul Moon.

UDIM reduces the loss landscape inconsistency between source domain and unknown domains. As unknown domains are inaccessible, these domains are empirically crafted by perturbing instances from the source domain dataset. In particular, by aligning the loss landscape acquired in the source domain to the loss landscape of perturbed domains, we expect to achieve generalization grounded on these flat minima for the unknown domains.

Our implementation code is largely dependent on the code of SAGM (CVPR2023). We thank the authors for providing these codes.

To reproduce the results of UDIM, we provide a bash file, which located at:

/bash/udim_pacs_xxx.sh

Here, XXX is a specific experimental setting.

This source code is released under the MIT license, included here. This project includes some code from DomainBed, also MIT licensed.

For additional questions or discussions, feel free to email me. ([email protected])