Open JonasMoros opened 1 year ago
Hi @JonasMoros, the MeshroomCL "MultiviewStereoCL" node that creates depth maps does not require a CUDA GPU, but it does require that you have one or more OpenCL-compatible devices installed. For most users, this is usually a GPU (from AMD, Intel, or NVIDIA), but can also be a CPU-based implementation of OpenCL.
This depth map creation stage of MeshroomCL is very compute-intensive, and has been optimized for use on a many-core OpenCL device such as a GPU. So while it will work on a CPU, it might take an impractically long amount of time. So if you don't have any kind of GPU, you might be better off using photogrammetry software that was designed with CPU-only processing in mind.
I understand that the normal version of Meshroom requires one to have a NIVIDIA GPU that has CUDA on it in order to run Depthmap. Does this version require it also or can I use my CPU instead?