You need a nvidia GPU with CUDA 9
installed. Note that CUDA 10
is not compatible with PCL 1.8
and doesn' work. If you want to build with CUDA 10
, please follow this document.
To choose the CUDA architecture compatible with your CUDA
version and graphics card, thereby also reducing compile time, pass -DCUDA_ARCH=<arg>
to CMake, where <arg>
is a two-digit compile compatibility version, e.g. "61", or choose this number from the corresponding drop-down in cmake-gui
.
In addition to CUDA
, the code depends on PCL
, OpenCV
and GLFW
. On Ubuntu, you can run the following command to install these dependencies:
sudo apt-get install libpcl-dev libopencv-dev libglfw3 libglfw3-dev
Now you are ready to build (remember to add cmake arguments as necessary):
git clone https://github.com/weigao95/surfelwarp
cd surfelwarp
mkdir build && cd build
cmake ..
make
If the build is successful, you might continue with the run instruction.