Domain Generalized Stereo Matching with Uncertainty-guided Data Augmentation
By: Shuangli Du , Jing Wang , Minghua Zhao and more
Potential Business Impact:
Trains fake-image AI to handle real photos
State-of-the-art stereo matching (SM) models trained on synthetic data often fail to generalize to real data domains due to domain differences, such as color, illumination, contrast, and texture. To address this challenge, we leverage data augmentation to expand the training domain, encouraging the model to acquire robust cross-domain feature representations instead of domain-dependent shortcuts. This paper proposes an uncertainty-guided data augmentation (UgDA) method, which argues that the image statistics in RGB space (mean and standard deviation) carry the domain characteristics. Thus, samples in unseen domains can be generated by properly perturbing these statistics. Furthermore, to simulate more potential domains, Gaussian distributions founded on batch-level statistics are poposed to model the unceratinty of perturbation direction and intensity. Additionally, we further enforce feature consistency between original and augmented data for the same scene, encouraging the model to learn structure aware, shortcuts-invariant feature representations. Our approach is simple, architecture-agnostic, and can be integrated into any SM networks. Extensive experiments on several challenging benchmarks have demonstrated that our method can significantly improve the generalization performance of existing SM networks.
Similar Papers
Class-invariant Test-Time Augmentation for Domain Generalization
CV and Pattern Recognition
Makes AI work better with new, unseen pictures.
Semantic-aware Random Convolution and Source Matching for Domain Generalization in Medical Image Segmentation
CV and Pattern Recognition
Helps doctors see inside bodies better with different scans.
TRUST: Leveraging Text Robustness for Unsupervised Domain Adaptation
CV and Pattern Recognition
Helps computers see better in new places.