Adaptive Multi-Modal Cross-Entropy Loss for Stereo Matching

1 code implementation • 27 Jun 2023 • Peng Xu, Zhiyu Xiang, Chenyu Qiao, Jingyun Fu, Tianyu Pu

Existing works simply assume that the ground-truth distributions are uni-modal, which ignores the fact that most of the edge pixels can be multi-modal.

Domain Generalization Stereo Matching