KPConv-PyTorch

Find a file

HuguesTHOMAS f4da047017 Corrections		2020-04-24 12:19:12 -04:00
cpp_wrappers	Initial commit	2020-03-31 15:42:35 -04:00
datasets	Corrections	2020-04-24 12:19:12 -04:00
kernels	Initial commit	2020-03-31 15:42:35 -04:00
models	Corrections	2020-04-24 12:00:11 -04:00
utils	Corrections	2020-04-24 12:00:11 -04:00
plot_convergence.py	Corrections	2020-04-24 12:00:11 -04:00
README.md	Initial commit	2020-03-31 15:56:08 -04:00
test_models.py	Initial commit	2020-04-11 15:15:46 -04:00
train_ModelNet40.py	Initial commit	2020-03-31 15:47:53 -04:00
train_S3DIS.py	Corrections	2020-04-24 12:19:12 -04:00
train_SemanticKitti.py	Corrections	2020-04-23 09:51:16 -04:00
visualize_deformations.py	Corrections	2020-04-24 12:00:11 -04:00

README.md

Created by Hugues THOMAS

Introduction

This repository contains the implementation of Kernel Point Convolution (KPConv) in PyTorch. [Work in progress]

KPConv is also available in Tensorflow (original implementation)

KPConv is a point convolution operator presented in our ICCV2019 paper (arXiv). If you find our work useful in your research, please consider citing:

@article{thomas2019KPConv,
    Author = {Thomas, Hugues and Qi, Charles R. and Deschaud, Jean-Emmanuel and Marcotegui, Beatriz and Goulette, Fran{\c{c}}ois and Guibas, Leonidas J.},
    Title = {KPConv: Flexible and Deformable Convolution for Point Clouds},
    Journal = {Proceedings of the IEEE International Conference on Computer Vision},
    Year = {2019}
}

Installation

TODO

Experiments

Currently only two experiments are available in Pytorch: classification on ModelNet40 and segmentation on S3DIS.

TODO: Guide for runnig experiments TODO: More experiments

Acknowledgment

Our code uses the nanoflann library.

License

Our code is released under MIT License (see LICENSE file for details).

Updates

31/03/2020: Initial release.