Uncertainty-guided Source-free Domain Adaptation (ECCV 2022)

Subhankar Roy, Martin Trapp, Andrea Pilzer, Juho Kannala, Nicu Sebe, Elisa Ricci and Arno Solin

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.

MAP vs Uncertainty-guided Target Adaptation

Pipeline of Uncertainty-guided Source-free Domain Adaptation (U-SFAN)

Installation

pip install -r requirements.txt

Toy-classification problem

cd toy_classification/

Mild-shift

SHOT-IM python multiclass_shot.py --method shot

U-SFAN python multiclass_shot.py --method usfan

Strong-shift

SHOT-IM python multiclass_shot.py --method shot --strong_shift

U-SFAN python multiclass_shot.py --method usfan --strong_shift

Image Datasets

CIFAR9 → STL9

cd cifar/

Source Model training + Bayesian hypothesis generation:

python main_cda.py --s CIFAR9 --t STL9 --download --gpu 0 --mode source

SHOT-IM (using MAP) Target Adaptation:

python main_cda.py --s CIFAR9 --t STL9 --gpu 0 --mode target --method shot --lr 1e-3

U-SFAN (using Bayesian hypothesis) Target Adaptation:

python main_cda.py --s CIFAR9 --t STL9 --gpu 0 --mode target --method usfan --lr 1e-3