Confidence-based and sample-reweighted test-time adaptation

Hao Yang, Min Wang, Zhengfei Yu, Hang Zhang, Jinshen Jiang, Yun Zhou

Abstract: Highlights•The first study focuses on the importance of samples in Test Time Adaptation.•We propose a confidence-based optimization strategy for parameter updating.•We propose a novel truncation function that assigns an adaptive weight to each sample.•Three digital-world and one physical-world datasets demonstrate the effectiveness of the proposed method.