Skip to content
master
Switch branches/tags
Code

Latest commit

 

Git stats

Files

Permalink
Failed to load latest commit information.
Type
Name
Latest commit message
Commit time
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

UKPGAN: Unsupervised KeyPoint GANeration.

This repo is a TensorFlow implementation of our work UKPGAN. UKPGAN is an unsupervised 3D keypoint detector where keypoints are detected so that they could reconstruct the original object shape. Note that our keypoint detector solely depends on local features and is both translational and rotational invariant.

intro

Quick Start

  1. Install Pybind11 and PCL C++ dependencies.
  2. Create env from environment.yml.
conda env create -f environment.yml
  1. Download ShapeNet point clouds from KeypointNet and unzip pcds folder to the root.

  2. Compile smoothed density value (SDV) source files.

cd sdv_src
mkdir build
cd build
cmake .. -DCMAKE_BUILD_TYPE=Release
make
cd ../..
  1. Open a separate terminal to monitor training process
visdom -port 1080
  1. Run
python train.py

Related Projects

Citation

If you find our algorithm useful in your research, please consider citing:

@article{you2020ukpgan,
  title={UKPGAN: Unsupervised KeyPoint GANeration},
  author={You, Yang and Liu, Wenhai and Li, Yong-Lu and Wang, Weiming and Lu, Cewu},
  journal={arXiv preprint arXiv:2011.11974},
  year={2020}
}