MCPDepth cuts mean absolute error by 18.8% on Deep360 and 19.9% on 3D60, using standard layers with a circular attention module for cylindrical panoramas. Read more: https://getnews.me/mcpdepth-better-omnidirectional-depth-with-multi-cylindrical-stereo/ #mcpdepth #cylindrical

MCPDepth, a two‑stage framework for omnidirectional depth, reduces mean absolute error by 18.8% on the Deep360 benchmark and 19.9% on the 3D60 dataset. https://getnews.me/mcpdepth-boosts-omnidirectional-depth-with-multi-cylindrical-stereo/ #mcpdepth #depthestimation

Feng Qiao, Zhexiao Xiong, Xinge Zhu, Yuexin Ma, Qiumeng He, Nathan Jacobs
MCPDepth: Omnidirectional Depth Estimation via Stereo Matching from Multi-Cylindrical Panoramas
https://arxiv.org/abs/2408.01653