这个 setting 的目的就是只利用源模型来完成 Domain Adaptation。 Test-Time Training (TTT):从信息的角度,从前我们训练神经网络都只利用了训练集的信息(监督学习),但其实测试集也从数据分布的角度提供了信息。这个 setting 主要就是提出了一种同时利用了训练集信息,和测试集所提供的数据分布的信息去训练神经网络的方...
transfer-learningdomain-adaptationtest-time-augmentationdomain-generalizationdistribution-shifttest-time-adaptationsource-free-domain-adaptationtest-time-trainingcontinual-test-time-adaptation UpdatedJan 17, 2025 YuejiangLIU/awesome-source-free-test-time-adaptation ...
transfer-learning domain-adaptation test-time-augmentation domain-generalization distribution-shift test-time-adaptation source-free-domain-adaptation test-time-training continual-test-time-adaptation Updated Feb 13, 2025 liangchen527 / ITTA Star 33 Code Issues Pull requests Official code for the CV...
To overcome this limitation, we propose a novel test-time adaptation method, called Test-time Adaptation via Self-Training with nearest neighbor information (TAST), which is composed of the following procedures: (1) adds trainable adaptation modules on top of the trained feature extractor; (2) ...
In recent years, self-training has also shown promising results by iteratively using gradually-improving target pseudo-labels to train the network [19, 36, 62, 75]. 2.2. Test-time Adaptation Test-time adaptation is also referred to as source-free do- main adaptation in some references [28,...
Testtime training引入了一种自监督辅助旋转预测任务,在源训练和目标训练中共同优化。但这种方法需要改变源训练协议,并非对所有模型都可行。此外,对比学习范式已被证明比作为文本前任务的旋转预测学习更多的可转移表征。最近,王德全等人的On-target Adaptation在预训练阶段使用了自监督学习,然而,作者认为这种方法并没有...
In this paper, we propose Test-Time Training, a general approach for improving the performance of predictive models when training and test data come from different distributions. We turn a single unlabeled test sample into a self-supervised learning problem, on which we update the model parameters...
Ding, Y., Sheng, L., Liang, J., Zheng, A., & He, R. (2023). Proxymix: Proxy-based mixup training with label refinery for source-free domain adaptation.Neural Networks,167, 92–103. Google Scholar D’Innocente, A., Borlino, F. C., Bucci, S., Caputo, B., & Tommasi, T. (...
Parameter-free Online Test-time Adaptation Malik Boudiaf E´ TS Montreal * Romain Mueller FiveAI Ismail Ben Ayed E´ TS Montreal Luca Bertinetto FiveAI Abstract Training state-of-the-art vision models has become pro- hibitively expensive for researchers and practitioners. For the sake of ...
First, we design a training-free dynamic adapter (TDA) that can achieve test-time adaptation of vision-language models efficiently and effectively. To the best of our knowledge, this is the first work that investigates the efficiency issue of test-time adaptation of vision-language models. Seco...