[ICME 2022] (Oral) ZSPU: "Zero-Shot" Point Cloud Upsampling

News

Self-training code released. PU and PS xyz data released. PU mesh data released.

Introduction

This repository is for our submitted paper for ICME 2022. The code is modified from PU-GAN, 3PU and PU-Net.

Installation

This repository is based on Tensorflow and the TF operators from PointNet++. Therefore, you need to install tensorflow and compile the TF operators.

For installing tensorflow, please follow the official instructions in here. The code is tested under TF1.15 (lower version should also work) with CUDA 10.0 and Python 3.6 on Ubuntu 18.04.

For compiling TF operators, please check tf_xxx_compile.sh or tf_xxx_compile_abi.sh under each op subfolder in code/tf_ops folder. Note that you need to update nvcc, cuda and tensoflow include library if necessary.

Note

When running the code, if you have undefined symbol: _ZTIN10tensorflow8OpKernelE error, you need to compile the TF operators. If you have already added the -I$TF_INC/external/nsync/public -L$TF_LIB -ltensorflow_framework but still have cannot find -ltensorflow_framework error. Please use 'locate tensorflow_framework ' to locate the tensorflow_framework library and make sure this path is in $TF_LIB.

Usage

Compile the TF operators Follow the above information to compile the TF operators.
Train the model: Test point clouds x are provided in folder data/[dataset]. Run:
```
python main.py --phase train --data_file x --use_data y
```
Here, x and y are the filename (without extension) for the point cloud and the index for dataset (0 for Data PU, 1 for Data Princeton, 2 for KITTI), respectively.
The prediction will be conducted automatically after the model is trained. You will see the input and output results in the folder data/[dataset]/input and data/[dataset]/output.
We will release all mesh files after the review period.

Evaluation code

We provide the code to calculate the uniform metric in the evaluation code folder. In order to use it, you need to install the CGAL library. Please refer this link and PU-Net to install this library. Then:

cd evaluation_code
cmake .
make
./evaluation ps/m0.off ps/m0.xyz

in where, the second argument is the mesh, and the third one is the predicted points.

Then, use evaluate.py to gain the quantitative results: python evaluate.py --pred ./evaluation_code/[dataset]/ --gt ./evaluation_code/[dataset]/gt/

Citation

Please consider citing this paper with the following bibtex if you are interested in this work:

@inproceedings{zhou2022zero,
  title={“Zero-Shot” Point Cloud Upsampling},
  author={Zhou, Kaiyue and Dong, Ming and Arslanturk, Suzan},
  booktitle={2022 IEEE International Conference on Multimedia and Expo (ICME)},
  pages={1--6},
  year={2022},
  organization={IEEE}
}

ky-zhou / ZSPU