Distribution alignment for fully test-time adaptation with dynamic online data streams
Given a model trained on source data, Test-Time Adaptation (TTA) enables adaptation and
inference in test data streams with domain shifts from the source. Current methods …
inference in test data streams with domain shifts from the source. Current methods …
How to train the teacher model for effective knowledge distillation
Recently, it was shown that the role of the teacher in knowledge distillation (KD) is to provide
the student with an estimate of the true Bayes conditional probability density (BCPD) …
the student with an estimate of the true Bayes conditional probability density (BCPD) …
[PDF][PDF] Optimal graph learning and nuclear norm maximization for deep cross-domain robust label propagation
Abstract Domain adaptation aims to achieve label transfer from a labeled source domain to
an unlabeled target domain, where the two domains exhibit different distributions. Existing …
an unlabeled target domain, where the two domains exhibit different distributions. Existing …
Adapting to Distribution Shift by Visual Domain Prompt Generation
In this paper, we aim to adapt a model at test-time using a few unlabeled data to address
distribution shifts. To tackle the challenges of extracting domain knowledge from a limited …
distribution shifts. To tackle the challenges of extracting domain knowledge from a limited …
Beyond model adaptation at test time: A survey
Machine learning algorithms have achieved remarkable success across various disciplines,
use cases and applications, under the prevailing assumption that training and test samples …
use cases and applications, under the prevailing assumption that training and test samples …
CovRoot: COVID-19 detection based on chest radiology imaging techniques using deep learning
The world first came to know the existence of COVID-19 (SARS-CoV-2) in December 2019.
Initially, doctors struggled to diagnose the increasing number of patients due to less …
Initially, doctors struggled to diagnose the increasing number of patients due to less …
Rethinking Normalization Layers for Domain Generalizable Person Re-identification
R Nie, J Ding, X Zhou, X Li - European Conference on Computer Vision, 2024 - Springer
Abstract Domain Generalizable Person Re-Identification (DG-ReID) strives to transfer
learned feature representation from source domains to unseen target domains, despite …
learned feature representation from source domains to unseen target domains, despite …
DocTTT: Test-Time Training for Handwritten Document Recognition Using Meta-Auxiliary Learning
W Gu, L Gu, Z Wang, CY Suen, Y Wang - arXiv preprint arXiv:2501.12898, 2025 - arxiv.org
Despite recent significant advancements in Handwritten Document Recognition (HDR), the
efficient and accurate recognition of text against complex backgrounds, diverse handwriting …
efficient and accurate recognition of text against complex backgrounds, diverse handwriting …
[PDF][PDF] Multi-attention based visual-semantic interaction for few-shot learning
Abstract Few-Shot Learning (FSL) aims to train a model that can generalize to recognize
new classes, with each new class having only very limited training samples. Since extracting …
new classes, with each new class having only very limited training samples. Since extracting …
Ddfp: Data-dependent frequency prompt for source free domain adaptation of medical image segmentation
Abstract Domain adaptation aims to address model performance degradation problem
under domain gap. In the typical setting of unsupervised domain adaptation, labeled data …
under domain gap. In the typical setting of unsupervised domain adaptation, labeled data …