Skip to content

Latest commit

 

History

History
47 lines (37 loc) · 1.96 KB

README.md

File metadata and controls

47 lines (37 loc) · 1.96 KB
Raw

Learning Symmetric Shapes

In this work, we propose an efficient iterative planar parameterization for disk topology shapes. The parameterization is used as a tool to regularize the mesh onto a square grid and encoded with vertex position. The resultant encoding is an image with rgb denoting xyz positions on the mesh.

The ICCV workshop paper can be found here. First part of this project deals with parameterization and second deals with learning of shapes from geometry images.

teaser

Installation

Code for Parameterization has been written in C++ and requires:

Deep network code is based on Tensorflow and is tested on Ubuntu with:

  • python (3.5.2)
  • tensorflow-gpu (1.14)
  • scikit-image (0.15.0)
  • numpy (1.16.5)
  • natsort
  • tqdm

Usage

Part I: Parameterization

Code contains functionality for:

  • slicing the mesh (--slice)
  • Iterative Surface Parameterization with n iterations (--sPI n)
  • Compute Geometry Image (of size im) from the parameterized representation (--m2G im)
  • Remesh point cloud from Geometry Image (--G2o)

Part II: Learning Shapes

python based functionality which contains:

  • generating curvature mask from normalGI
  • tensorflow model
  • docker image
  • python scripts to train and test the model

Citation

If you find this project useful in your work, please consider citing:

@inproceedings{jain2019learning,
  title={Learning to Reconstruct Symmetric Shapes using Planar Parameterization of 3D Surface},
  author={Jain, Hardik and Wöllhaf, Manuel and Hellwich, Olaf},
  booktitle={The IEEE International Conference on Computer Vision (ICCV) Workshops},
  year={2019}
}