PCRNet: Point Cloud Registration Network using PointNet Encoding

Source Code Author: Vinit Sarode and Xueqian Li

Paper | Website | Video | Pytorch Implementation

Requirements:

1. Cuda 10
2. tensorflow==1.14
3. transforms3d==0.3.1
4. h5py==2.9.0

Dataset:

Path for dataset: Link

1. Download 'train_data' folder from above link for iterative PCRNet.
2. Download 'car_data' folder from above link for PCRNet.

Pretrained Model:

Download pretrained models from Link

How to use code:

Compile loss functions:

1. cd utils/pc_distance
2. make -f makefile_10.0 clean
3. make -f makefile_10.0

Train Iterative-PCRNet:

1. chmod +x train_itrPCRNet.sh
2. ./train_itrPCRNet.sh

Train PCRNet:

1. chmod +x train_PCRNet.sh
2. ./train_PCRNet.sh

Citation

@InProceedings{vsarode2019pcrnet,
       author = {Sarode, Vinit and Li, Xueqian and Goforth, Hunter and Aoki, Yasuhiro and Arun Srivatsan, Rangaprasad and Lucey, Simon and Choset, Howie},
       title = {PCRNet: Point Cloud Registration Network using PointNet Encoding},
       month = {Aug},
       year = {2019}
}