1 | initial version |
Long initialization time is a known issue. The gpu module is focused on long time processing (e.g. video processing), where initialization time is not important. The best way is to reorganize your application so it processes several inputs at once (e.g. several templates). Also to reduce initialization time you can build OpenCV gpu module only for your card's compute capability. Set CMake variables CUDA_ARCH_BIN="1.3"
and CUDA_ARCH_PTX=""
.