Nicholas Lawrance

Scalable and Differentiable Point-Cloud Registration Using Maximum Mean Discrepancy

Rixon Crane, Fahira Afzal Maken, Nicholas Lawrance, Stanislav Funiak, Kasra Khosoussi, Ming Xu, et al.

We present MMD-Reg, a novel correspondence-free approach to point-cloud registration that is differentiable and has linear computational complexity in the number of points. We model registration as a nonlinear least-squares problem based on the Maximum Mean Discrepancy, approxima…Read more