VPC-Net: Completion of 3D Vehicles from MLS Point Clouds ISPRS Journal (IF = 7.319)
Yan Xia, Yusheng Xu, Cheng Wang, Uwe Stilla
Technical University of Munich, Xiamen University
VPC-Net is a neural network to synthesize complete, dense, and uniform point clouds for vehicles from MLS data. The arXiv version of VPC-Net can be found here.
@article{xia2021vpc, title={VPC-Net: Completion of 3D vehicles from MLS point clouds}, author={Xia, Yan and Xu, Yusheng and Wang, Cheng and Stilla, Uwe}, journal={ISPRS Journal of Photogrammetry and Remote Sensing}, volume={174}, pages={166--181}, year={2021}, publisher={Elsevier} }
This code is built using Tensorflow 1.12 with CUDA 9.0 and tested on Ubuntu 16.04 with Python 3.5.
Please follow PointNet++ to compile TF operators.
- Download
shapenet_car
directory on Google Drive. - Run
python train.py
- ShapenNet-car completion
- run
python test.py
.
- run
- KITTI completion
- Download KITTI data from PCN project.
- run
python test_kitti.py
.
Our implementations based on PCN repository.