A Novel Local Geometry Capture in PointNet++ for 3D Classification
Shivanand Venkanna Sheshappanavar, Chandra Kambhamettu; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2020, pp. 262-263
Abstract
Few of the recent deep learning models for 3D point sets classification are dependent on how well the model captures the local geometric structures. PointNet++ model made remarkable progress in learning local geometric structures than its predecessor PointNet. It recursively applies PointNet on nested partitions of the input 3D point set. PointNet++ model was able to extract the local region features from points by ball querying the local neighborhoods. However, ball querying is less effective in capturing local neighborhoods of high curvature surfaces or regions. In this paper, we demonstrate improvement in the 3D classification results by using ellipsoid querying around centroids, capturing more points in the local neighborhood. We extend the ellipsoid querying technique by orienting it in the direction of principal axes of the local neighborhood for better capture of the local geometry. We then take the union of points grouped by ball querying and ellipsoid querying with re-orientation to improve the PointNet++ classification results by 1.1%. Furthermore, we demonstrate the impact of re-oriented ellipsoid querying on a state-of-the-art ball query-based model, Relation-Shape Convolutional Neural Network (RS-CNN), with a 0.8% improvement in classification accuracy on ModelNet40 dataset.
Related Material
[pdf]
[
bibtex]
@InProceedings{Sheshappanavar_2020_CVPR_Workshops,
author = {Sheshappanavar, Shivanand Venkanna and Kambhamettu, Chandra},
title = {A Novel Local Geometry Capture in PointNet++ for 3D Classification},
booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
month = {June},
year = {2020}
}