KAIST-VCLAB / texturefusion

[CVPR2020] TextureFusion: High-Quality Texture Acquisition for Real-Time RGB-D Scanning
Other
121 stars 21 forks source link

CUDACameraTrackingMultiResRGB.cu is missing #9

Open grungeguitarist opened 2 years ago

grungeguitarist commented 2 years ago

CUDACameraTrackingMultiResRGB.cu is missing and thus renderCorrespondenceCUDA() is not working.

However, the file from NormalFusion is working: https://github.com/KAIST-VCLAB/normal-fusion/blob/master/Source/CUDACameraTrackingMultiResRGB.cu

grungeguitarist commented 2 years ago

BTW, here is a step-by-step installation that worked for me...:

  1. Make original VoxelHashing running


2. Insert TextureFusion

- Download from https://github.com/KAIST-VCLAB/texturefusion
- Copy all files to the VoxelHashing root folder "DepthSensingCUDA" and overwrite all existing files
- Open .sln in Visual Studio 2013
- In VS2013, replace content of "DepthSensing.cpp" and "DepthSensing.h" with content of downloaded texFusion.cpp" and "DepthSensing.h", respectively
- In VS2013, rename "DepthSensing.cpp" to "texFusion.cpp" and "DepthSensing.h" to "texFusion.h"
- Drag and drop the following downloaded files to the root (DepthSensing) in the Solution Explorer of VS2013
    - GlobalWarpingState.cpp
    - GlobalWarpingState.h
    - modeDefine.h
    - texFusion_main.cpp
    - texturePool.h
    - TexUpdateUtil.h
- Drag and drop the following downloaded files to the "CUDA" dir in VS2013
    - CUDABuildLinearSystemLocalRGB.cpp
    - CUDABuildLinearSystemLocalRGB.cu
    - CUDABuildLinearSystemLocalRGB.h
    - CUDABuildLinearSystemRGB.cpp
    - CUDABuildLinearSystemRGB.cu
    - CUDABuildLinearSystemRGB.h
    - CUDACameraTrackingLocalRGB.cpp
    - CUDACameraTrackingLocalRGB.cu
    - CUDACameraTrackingLocalRGB.h
    - CUDACameraTrackingMultiResRGB.cpp
    - CUDACameraTrackingMultiResRGB.h
    - cudaDebug.h
    - CUDAImageUtil.cu
    - CUDAImageUtil.h
    - CUDATexPoolParams.h
    - CUDATexUpdate.cpp
    - CUDATexUpdate.cu
    - CUDATexUpdate.h
    - CUDATexUpdateParams.h
- Drag and drop the following downloaded files to the "Sensors" dir in VS2013
    - SensorDataReaderZhou.cpp
    - SensorDataReaderZhou.h
    - sensordatazhou.cpp
    - sensordatazhou.cpp
- One file is missing, copy it from the follow-up paper NormalFusion
    - Download CUDACameraTrackingMultiResRGB.cu from https://github.com/KAIST-VCLAB/normal-fusion/blob/master/Source/CUDACameraTrackingMultiResRGB.cu
    - Move the file to the "Source" dir
    - Drag and drop the file to the "CUDA" dir in VS2013

3. Install required libs
- Install CUDA 10.1
- Install OpenCV 3.4.1
    - Downlaoad from https://github.com/opencv/opencv/archive/3.4.1.zip
    - Open Cmake
        - source code: C:/opencv-3.4.1
        - build the binaries: C:/opencv-3.4.1/build
        - configure
        - choose Visual Studio 2013 (!)
        - BUILD_opencv_world: check
        - WITH_CUDA: check
        - Change path of CUDA_TOOLKIT_ROOT_DIR to "10.1" version if a different one was chosen by Cmake
        - Download Eigen 3.3.4 https://gitlab.com/libeigen/eigen/-/releases/3.3.4
        - Change path of EIGEN_INCLDE_PATH to the downloaded "3.3.4" version 
        - configure
        - generate
        - open project in VS 2013
        - Change to Release Mode
        - right click on INSTALL and "Build"

4. Link required libs
    - Right click on DepthSensing -> Build Dependencies -> Build Customizations...
        - Uncheck CUDA 7.5
        - Check CUDA 10.1
    - Right click on DepthSensing -> Proporties -> Configuration Proporties
    - Add "C:\opencv-3.4.1\build\install\include" to...
        - VC++ Directories -> Include Directories 
        - C/C++ -> Additional Include Directories
        - CUDA C/C++ -> Additional Include Directories
    - Add "C:\opencv-3.4.1\build\install\x64\vc12\lib" to...
        - Linker -> Additional Library Directories
    - Add "C:\opencv-3.4.1\build\install\x64\vc12\lib\opencv_world341.lib" to...
        - Linker -> Input -> Additional Dependencies
    - Add "C:\opencv-3.4.1\build\install\x64\vc12\bin" to...
        - Path of Environment Variables on Windows

5. Run
    - Change to Release Mode in VS2013
    - Download fountain data set from http://vclab.kaist.ac.kr/cvpr2020p1/fountain_all.zip
    - Put content to D:/dataset/fountain_all
    - Run in VS2013
    - Press 9 when reconstructing was finished to generate .obj file