Skip to content

Latest commit

 

History

History
39 lines (36 loc) · 1.18 KB

README.md

File metadata and controls

39 lines (36 loc) · 1.18 KB

Spatiotemporal Texture Reconstruction

This repository contains the official implementation of the following paper:

Spatiotemporal Texture Reconstruction for Dynamic Objects Usinga Single RGB-D Camera (Eurographics 2021) Teaser

Requirements

  • Windows 10
  • CUDA >= 10.2
  • python >= 3.6
  • MSVC++ >= 14.2
  • OpenCV >= 3.4
  • Eigen
  • OpenGL
  • jsoncpp
  • TBB
  • mapmap (https://github.com/dthuerck/mapmap_cpu)
    Libaries are in zip file. Please extract them before build program.

Quick Start

  1. Parameterization
    UVAtlas code: https://github.com/microsoft/UVAtlas
  2. Foreground extraction
    You can find foreground extraction code in https://github.com/csaishih/foreground-extraction
  3. Global texture coordinate optimization
    Build and execute TextureMappingNonRigid project.
  4. Prepare labeling
    In ./Similarity folder, python ./simchek.py python ./variance.py In SHOT folder, python ./shot.py
  5. Labeling
    Re-execute TextureMappingNonRigid project.
  6. Color correction
    We use Texture Stitching program: https://github.com/mkazhdan/TextureSignalProcessing
  7. View result
    In conf.json file, convert "is_viewer" false to true, and execute.