Adaptive Multi-Modal Cross-Entropy Loss for Stereo Matching

arXiv:2306.15612v2 Announce Type: replace
Abstract: Despite the great success of deep learning in stereo matching, recovering accurate disparity maps is still challenging. Currently, L1 and cross-entropy are the two most widely used losses for stereo network training. Compared with the former, the latter usually performs better thanks to its probability modeling and direct supervision to the cost volume. However, how to accurately model the stereo ground-truth for cross-entropy loss remains largely under-explored. Existing works simply assume that the ground-truth distributions …

abstract arxiv cross-entropy cs.cv deep learning entropy loss losses maps modal modeling multi-modal network network training probability success training type