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)

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

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