1 introduction iccv23 task: source-free domain generalization common practice of DG: Utilize multiple source domains for learning domain-invariant features, but it is not clear which source domains are ideal for DG. Furthermore, collecting such large-scale multi-source domain data is costly. anothe...
而对于source-domain的性能,文章采取了0.9/0.1的数据集分割策略,可以看到本文的方法在目标域的效果显著,同时在源域也有非常不错的性能。 最后,欢迎大家关注github,聚合了OOD,causality,robustness以及optimization的一些阅读笔记 https://github.com/yfzhang114/Generalization-Causalitygithub.com/yfzhang114/Generalizatio...
Source-free domain generalization (SFDG) tackles the challenge of adapting models to unseen target domains without access to source domain data. To deal with this challenging task, recent advances in SFDG have primarily focused on leveraging the text modality of vision-language models such as CLIP....
Despite having diverse features important for generalization, the pre-trained feature extractor can overfit to the source data distribution during source training and forget relevant target domain knowledge. Rather than discarding this valuable knowledge, we introduce an integrated framework to incorporate ...
Source-Free Unsupervised Domain Adaptation: Current research and future directions 2024, Neurocomputing Show abstract SAN-Net: Learning generalization to unseen sites for stroke lesion segmentation with self-adaptive normalization 2023, Computers in Biology and Medicine Citation Excerpt : For site-invariant...
Test-Time Training with Self-Supervision for Generalization under Distribution Shifts (ICML 2020) 动机: Distribution Shifts指的是训练和测试阶段图像的风格发生变化,有点类似于DA的source和target domain。但是与DA不同的是,TTT在测试时只有测试图像,可以根据无标注的测试样本更新模型。但是TTT一般的设置是只根据sam...
, under the OS-SFDA setting, where the target domain has a additional abundant class distribution than the source domain, mentioned OS-SFDA methods group unknown classes as one, which restricts the thorough ex- ploration of target class information, hampering knowled...
The proposed source-free domain adaptation is broadly divided into two: vendor-side and client-side. 3.1 Vendor-side Strategy In the absence of target data, the vendor’s task effectively reduces to domain generalization (DG) [40]. DG is shown to be highly effective in the presence of ...
Relevant Knowledge Replay Stage: This stage is dedicated to enhancing the model’s understanding and generalization ability of knowledge and alleviating the catastrophic forgetting problem. The multi-stage approach complements each ot...
to improve the target adaptation. We introduce consistency between label preserving augmentations and utilize pseudo label refinement methods to reduce noisy pseudo labels. Further, we propose novel MixUp Knowledge Distillation (MKD) for better generalization on multiple target domains using various source...