prstrive / UniMVSNet

[CVPR 2022] Rethinking Depth Estimation for Multi-View Stereo: A Unified Representation
MIT License
228 stars 12 forks source link

关于坦克和寺庙数据集的调参问题! #25

Open Wuuu3511 opened 1 year ago

Wuuu3511 commented 1 year ago

您好,首先感谢您优秀的工作! 我最近也在从事MVS相关工作,在阅读您的代码时发现,坦克和寺庙数据集每个场景的光度一致性滤波的参数各不相同,想请教一下,当使用dypcd时如何获取每个场景最优的参数。 感谢您的阅读与回复!

lt-xiang commented 1 year ago

同问

prstrive commented 1 year ago

We determined the parameters of each scene during the experiment of reproducing our baseline.