KPConv: Flexible and Deformable Convolution for Point Clouds - Archive ouverte HAL Access content directly
Conference Papers Year : 2019

KPConv: Flexible and Deformable Convolution for Point Clouds

Charles R. Qi
  • Function : Author
Jean-Emmanuel Deschaud
Francois Goulette
Leonidas J. Guibas
  • Function : Author
  • PersonId : 850076


We present Kernel Point Convolution (KPConv), a new design of point convolution, i.e. that operates on point clouds without any intermediate representation. The convolution weights of KPConv are located in Euclidean space by kernel points, and applied to the input points close to them. Its capacity to use any number of kernel points gives KPConv more flexibility than fixed grid convolutions. Furthermore, these locations are continuous in space and can be learned by the network. Therefore, KPConv can be extended to deformable convolutions that learn to adapt kernel points to local geometry. Thanks to a regular subsampling strategy, KPConv is also efficient and robust to varying densities. Whether they use deformable KPConv for complex tasks, or rigid KPconv for simpler tasks, our networks outperform state-of-the-art classification and segmentation approaches on several datasets. We also offer ablation studies and visualizations to provide understanding of what has been learned by KPConv and to validate the descriptive power of deformable KPConv.

Dates and versions

hal-02310026 , version 1 (09-10-2019)



Hugues Thomas, Charles R. Qi, Jean-Emmanuel Deschaud, Beatriz Marcotegui, Francois Goulette, et al.. KPConv: Flexible and Deformable Convolution for Point Clouds. The IEEE International Conference on Computer Vision (ICCV), Oct 2019, Séoul, South Korea. ⟨hal-02310026⟩
4806 View
0 Download



Gmail Facebook X LinkedIn More