Uncertainty-guided source-free domain adaptation

Loading...
Thumbnail Image
Journal Title
Journal ISSN
Volume Title
A4 Artikkeli konferenssijulkaisussa
Date
2022
Department
University of Trento
Department of Computer Science
Computer Science Professors
Major/Subject
Mcode
Degree programme
Language
en
Pages
537-555
Series
Computer Vision – ECCV 2022, Lecture Notes in Computer Science, Volume 13685
Abstract
Source-free domain adaptation (SFDA) aims to adapt a classifier to an unlabelled target data set by only using a pre-trained source model. However, the absence of the source data and the domain shift makes the predictions on the target data unreliable. We propose quantifying the uncertainty in the source model predictions and utilizing it to guide the target adaptation. For this, we construct a probabilistic source model by incorporating priors on the network parameters inducing a distribution over the model predictions. Uncertainties are estimated by employing a Laplace approximation and incorporated to identify target data points that do not lie in the source manifold and to down-weight them when maximizing the mutual information on the target data. Unlike recent works, our probabilistic treatment is computationally lightweight, decouples source training and target adaptation, and requires no specialized source training or changes of the model architecture. We show the advantages of uncertainty-guided SFDA over traditional SFDA in the closed-set and open-set settings and provide empirical evidence that our approach is more robust to strong domain shifts even without tuning.
Description
Keywords
Citation
Roy , S , Trapp , M , Pilzer , A , Kannala , J , Sebe , N , Ricci , E & Solin , A 2022 , Uncertainty-guided source-free domain adaptation . in S Avidan , G Brostow , M Cissé , G M Farinella & T Hassner (eds) , Computer Vision – ECCV 2022 : 17th European Conference, Tel Aviv, Israel, October 23–27, 2022, Proceedings, Part XXV . Lecture Notes in Computer Science , vol. 13685 , Springer , pp. 537-555 , European Conference on Computer Vision , Tel Aviv , Israel , 23/10/2022 . https://doi.org/10.1007/978-3-031-19806-9_31