search

chrischoy / DeepGlobalRegistration

[CVPR 2020 Oral] A differentiable framework for 3D registration
Other
477 stars 85 forks source link

Questions regarding the outlier rejection module. #4

Closed XuyangBai closed 4 years ago

XuyangBai commented 4 years ago

Hi Chris @chrischoy

Thanks for sharing this interesting work. I wonder have you tried the Context Normalization in the outlier filtering module, which is commonly used in 2D outlier rejection methods? And will using some pre-filtering strategies like mutual check or ratio test give better performance to your network?

Best, Xuyang.

chrischoy commented 4 years ago

Yes, I tried on the High-Dimensional Convolutional Networks for Geometric Pattern Recognition, CVPR'20 in a slightly different setup and found that it cannot be trained well with the 360 rotation augmentation, and leading to lower performance.