Project: pointnet-registration-framework (GitHub Link)

pointnet-registration-framework-master
- pcrnet
  - utils
    - data_prep_util.py
    - tf_util.pyc
    - __pycache__
      - tf_util.cpython-35.pyc
      - tf_util_loss.cpython-35.pyc
    - eulerangles.py
    - create_dataset
      - classes_test.txt
      - classes_train.txt
      - generate_poses.py
      - generate_dataset.py
      - single_catg.txt
      - noise_test.txt
    - pc_distance
      - tf_nndistance.cu.o
      - makefile_8.0
      - tf_approxmatch.pyc
      - __pycache__
        tf_approxmatch.cpython-35.pyc
        tf_nndistance.cpython-35.pyc
        __init__.cpython-35.pyc
      - tf_approxmatch.cu
      - __init__.pyc
      - tf_approxmatch_so.so
      - tf_nndistance.cu
      - tf_approxmatch.cpp
      - makefile_10.0
      - __init__.py
      - tf_nndistance.pyc
      - tf_nndistance.cpp
      - tf_nndistance_so.so
      - tf_nndistance.py
      - tf_approxmatch.py
      - tf_approxmatch.cu.o
    - tf_util_loss.py
    - pc_util.py
    - tf_util_loss.pyc
    - plyfile.py
    - tf_util.py
  - iterative_PCRNet.py
  - LICENSE
  - models
    - __pycache__
      - pointnet_pose.cpython-35.pyc
      - pcr_model.cpython-35.pyc
    - pcr_model.py
    - ipcr_model.py
  - visualize_results.py
  - results_itrPCRNet.py
  - train_PCRNet.sh
  - results_PCRNet.py
  - README.md
  - requirements.txt
  - plot_results
    - plot_threshold_vs_success_trans.py
    - labels.txt
    - test_icp_train_data_45_maxItr100_noise
      - test.txt
    - folders.txt
    - plot_threshold_vs_success.py
    - multi_catg_noise
    - test_PNLK_multi_catg_noise_train_data_noise
      - test.txt
      - params.txt
    - test_itr_multi_catg_noise_train_data_45
      - test.txt
      - params.txt
  - helper.py
  - iterative_PCRNet_partial.py
  - create_partial_data.py
  - train_itrPCRNet.sh
  - PCRNet.py
  - test_icp
    - analysis_files
      - icp.py
      - helper_analysis.pyc
      - __init__.pyc
      - __init__.py
      - helper_analysis.py
      - icp.pyc
    - test_icp.py
- images
- LICENSE
- README.md
- pnlk
  - experiments
    - test_icp_vs_pointlk.py
    - generate_rotations.py
    - ex1_genpert.sh
    - icp.py
    - test_utils.py
    - ex1_genrot.sh
    - ex1_result_stat.sh
    - ex1_test_pointlk.sh
    - ex1_test_icp.sh
    - .DS_Store
    - test_icp.py
    - ex1_train.sh
    - test_partial_data.py
    - sampledata
      - modelnet40_half2.txt
      - shapenet2_20.txt
      - modelnet40_half1.txt
      - modelnet40.txt
    - generate_perturbations.py
    - train_classifier.py
    - result_stat.py
    - train_pointlk.py
    - test_pointlk.py
  - ptlk
    - __pycache__
      - pointlk.cpython-36.pyc
      - pointnet.cpython-36.pyc
      - invmat.cpython-36.pyc
      - so3.cpython-36.pyc
      - __init__.cpython-36.pyc
      - se3.cpython-36.pyc
      - sinc.cpython-36.pyc
    - sinc.py
    - se3.py
    - so3.py
    - invmat.py
    - pointnet.py
    - data
      - mesh.py
      - __pycache__
        globset.cpython-36.pyc
        transforms.cpython-36.pyc
        __init__.cpython-36.pyc
        datasets.cpython-36.pyc
        mesh.cpython-36.pyc
      - datasets.py
      - __init__.py
      - globset.py
      - transforms.py
    - pointlk.py
    - __init__.py

One Framework to Register Them All: PointNet Encoding for Point Cloud Alignment

Source Code Author: Vinit Sarode, Xueqian Li and Animesh Dhagat

Requirements:

Cuda 10
tensorflow==1.14
transforms3d==0.3.1
h5py==2.9.0
pytorch==1.3.0

Dataset:

Path for dataset: Link

Download 'train_data' folder from above link.
Download 'car_data' folder from above link.

Point Cloud Registration Network:

Train Iterative-PCRNet:

cd pcrnet
chmod +x train_itrPCRNet.sh
./train_itrPCRNet.sh

Note: PCRNet's pytorch implementation is also available at [Link].

PointNetLK:

Train PointNetLK:

cd pnlk
cd experiments
python train_pointlk.py

Citation

@misc{sarode2019framework,
    title={One Framework to Register Them All: PointNet Encoding for Point Cloud Alignment},
    author={Vinit Sarode and Xueqian Li and Hunter Goforth and Yasuhiro Aoki and Animesh Dhagat and Rangaprasad Arun Srivatsan and Simon Lucey and Howie Choset},
    year={2019},
    eprint={1912.05766},
    archivePrefix={arXiv},
    primaryClass={cs.CV}
}