3D Point Cloud Registration with Multi-Scale Architecture and Unsupervised Transfer Learning
This addresses the challenge of applying point cloud registration methods to unseen datasets without labeled data, which is incremental as it builds on existing deep learning approaches.
The paper tackles the problem of generalizing deep learning for 3D point cloud registration to new datasets by combining a multi-scale neural network (MS-SVConv) with an unsupervised transfer learning algorithm (UDGE), achieving state-of-the-art results on real-world datasets like 3DMatch, ETH, and TUM.
We propose a method for generalizing deep learning for 3D point cloud registration on new, totally different datasets. It is based on two components, MS-SVConv and UDGE. Using Multi-Scale Sparse Voxel Convolution, MS-SVConv is a fast deep neural network that outputs the descriptors from point clouds for 3D registration between two scenes. UDGE is an algorithm for transferring deep networks on unknown datasets in a unsupervised way. The interest of the proposed method appears while using the two components, MS-SVConv and UDGE, together as a whole, which leads to state-of-the-art results on real world registration datasets such as 3DMatch, ETH and TUM. The code is publicly available at https://github.com/humanpose1/MS-SVConv .