LSANet: Feature Learning on Point Sets by Local Spatial Aware Layer
The official implementation of LSANet in tensorflow.
Thanks to Xuan-Yi Li,and Deng-Ping Fan's help.
Introduction
We propose a new network layer, named Local Spatial Aware (LSA) Layer, to model geometric structure
in local region accurately and robustly. Each feature extracting operation in LSA layer is related to Spatial Distribution Weights (SDWs), which are learned based on the spatial distribution in local region, to establish a strong link with inherent geometric shape. The experiments show that our LSA-based network, named LSANet, can achieve on par or better performance than the state-of-the-art methods when evaluating on the challenging benchmark datasets. The network architecture of LSANet and LSA module are shown below.
Installation
The code is based on PointNet++. Please follow the instruction in PointNet++ to compile the customized TF operators.
Usage
Classification
ModelNet40 dataset can be downloaded here.
To train a LSANet for classification, please run
python train_multi_gpu.py
Part segmentation
ShapeNet dataset can be downloaded here. To train a LSANet for segmentation, please run
cd part_seg
python train_multi_gpu_one_hot.py
Useful links
Awesome point set learning
Citation
Please cite this paper if it is helpful to your research,
@article{chen2019lsanet,
title={LSANet: Feature Learning on Point Sets by Local Spatial Aware Layer},
author={Chen, Lin-Zhuo and Li, Xuan-Yi and Fan, Deng-Ping and Cheng, Ming-Ming and Wang, Kai and Lu, Shao-Ping},
journal={arXiv preprint arXiv:1905.05442},
year={2019}
}