Tested on ubuntu18.04 + python3.6 + cuda10.2 and win10 + python3.8 + cuda11
pip3 install numpy opencv-python open3d pycuda numba Cython
-
安装Python3.8, 建议 Python3.8.6 or Python3.8.10，安装时注意选择把python添加进path
-
下载安装VS2019 (Linux跳过此步): https://visualstudio.microsoft.com/zh-hans/vs/community/ 安装VS2019完毕后需要添加"cl.exe"的环境变量，示例:
C:\Program Files (x86)\Microsoft Visual Studio\2019\Community\VC\Tools\MSVC\14.29.30133\bin\Hostx64\x64 -
下载安装CUDA11.4：https://developer.nvidia.com/cuda-downloads 注意更换显卡或CUDA版本后需要将*.cubin文件删除，下次运行时会自动重新编译
-
通过Pip安装packages
pip config set global.index-url https://pypi.tuna.tsinghua.edu.cn/simple pip install numpy opencv-python pyserial hidapi open3d pycuda -
输出编译好的库: 把x3d_camera放置在跟本repo相同的目录下，然后运行
python setup.py build_ext --inplace会输出编译好的库到x3d_camera，Win下是pyd文件，linux下是so文件
Windows:
python.exe structured_light.py pattern_examples\struli_test1\
python.exe structured_light_cuda.py pattern_examples\struli_test1\
Linux:
python3 structured_light.py pattern_examples/struli_test1/
python3 structured_light_cuda.py pattern_examples/struli_test1/
Result is in ".\pattern_examples\struli_test1\res". You can use meshlab to visualize the ply file. https://www.meshlab.net/#download
When runing for the first time or src code is modified, it will need more time for numba or cuda compiling.
To make running faster when using CPU, set option use_parallel_computing to True in "structured_light.py". Note this may cause numba llvm compiling fails for some version of numba.