Efficient Point Cloud Upsampling and Normal Estimation using Deep Learning for Robust Surface Reconstruction

To run the project add root folder of the project to python path.export PYTHONPATH="ROOTPATH_OF_PROJECT:$PYTHONPATH" e.g., export PYTHONPATH="/home/user/point-normals-upsampling:$PYTHONPATH"

Setup

• Use anaconda for python3.7. Install requirements.txt. Install torch and cuda toolkit conda install pytorch torchvision cudatoolkit=10.1 -c pytorch
• Build pointnet++ module run python setup.py build_ext --inplace in root folder of project
• Build sampling module run python setup.py install in sampling folder of project
• Add absolute path of chamfer_distace.cpp and chamfer_distance.cu in chamfer_distance.py

Training

• For this repo we used PU-NET dataset for training. Download the hdf5 format patches dataset from GoogleDrive
• For training and evalutation run all commands inside code folder.
• Training: python train.py --num_points 1024 --checkpoint_path .. --batch_size 20 --epochs 400 --h5_data_file dataset_path e.g., python train.py --num_points 1024 --checkpoint_path .. --batch_size 20 --epochs 400 --h5_data_file ../data.h5
• Evaluation: python evaluate.py --test_file filename(.xyz) --num_points num (default=1024) --patch_num_ratio num (default=4) --trained_model checkpoint_path e.g., python evaluate.py --test_file ../test.xyz --num_points 1024 --patch_num_ratio 4 --trained_model ../checkpoint
• All the results will be saved in results folder in root directory

Acknowledgement

• PointNet++ PyTorch Implementation: erikwijmans/Pointnet2_PyTorch
• Official PyTorch: charlesq34/pointnet2
• PyTorch Chamfer Distance: chrdiller/pyTorchChamferDistance
• Patch-base progressive 3D Point Set Upsampling: yifita/3PU_pytorch