Uncertainty-guided Source-free Domain Adaptation
This work addresses the challenge of adapting classifiers to unlabeled target data without access to source data, which is incremental as it builds on existing SFDA methods by incorporating uncertainty estimation.
The paper tackles the problem of unreliable predictions in source-free domain adaptation (SFDA) due to domain shift and absence of source data, by proposing a method that quantifies uncertainty in source model predictions to guide target adaptation, resulting in improved robustness to strong domain shifts without tuning, as shown in closed-set and open-set settings.
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.