I'm trying to compile OpenCV 2.4.13 from source with CUDA 9.0, CUDNN 7.0,as suggested by the repo called "MonocularTotalCapture" that you can find here :
https://github.com/CMU-Perceptual-Computing-Lab/MonocularTotalCapture
I've installed Ubuntu 16.04 on a GTX 1060 GPU machine. I installed cuda 9 and CUDNN 7. Now I'm trying to install OpenCV 2.4.13,but I'm getting CMake Error:
CMake Error: The following variables are used in this project, but they are set to NOTFOUND.
Please set them or make sure they are set and tested correctly in the CMake files:
CUDA_nppi_LIBRARY (ADVANCED)
And then a very long list of targets like so:
linked by target "opencv_core" in directory /home/ziom/Scrivania/MonocularTotalCapture/opencv-2.4.13.6/modules/core
I know that a question like mine has been already solved here :
CMake Error: Variables are set to NOTFOUND
and I tried to follow the suggestions given there,but I haven't been able to fix my isse,maybe because I have a little different scenario. Anyway,I followed the suggestion of @A. B. M. Nazibullah :
Replacing FindCUDA.cmake and OpenCVDetectCUDA.cmake from https://github.com/opencv/opencv/tree/master/cmake (master branch- opencv 3.4.1) works for me for opencv_2.4.13 with NVIDIA TitanXP graphics card of Pascal Architecture and with cuda-9.0 on Ubuntu 16.04.
Infact,the first file to modify,called "FindCUDA.cmake" has been modified correctly,according with the suggestion of @api55. Infact this piece of code has been added at the place of the old one :
if(CUDA_VERSION VERSION_GREATER "5.0")
# In CUDA 5.5 NPP was splitted onto 3 separate libraries.
find_cuda_helper_libs(nppc)
#find_cuda_helper_libs(nppi)
find_cuda_helper_libs(nppial)
find_cuda_helper_libs(nppicc)
find_cuda_helper_libs(nppicom)
find_cuda_helper_libs(nppidei)
find_cuda_helper_libs(nppif)
find_cuda_helper_libs(nppig)
find_cuda_helper_libs(nppim)
find_cuda_helper_libs(nppist)
find_cuda_helper_libs(nppisu)
find_cuda_helper_libs(nppitc)
find_cuda_helper_libs(npps)
#set(CUDA_npp_LIBRARY "${CUDA_nppc_LIBRARY};${CUDA_nppi_LIBRARY};${CUDA_npps_LIBRARY}")
set(CUDA_npp_LIBRARY "${CUDA_nppc_LIBRARY};${CUDA_nppial_LIBRARY};${CUDA_nppicc_LIBRARY};${CUDA_nppicom_LIBRARY};${CUDA_nppidei_LIBRARY};${CUDA_nppif_LIBRARY};${CUDA_nppig$
elseif(NOT CUDA_VERSION VERSION_LESS "4.0")
find_cuda_helper_libs(npp)
endif()
if(NOT CUDA_VERSION VERSION_LESS "7.0")
# cusolver showed up in version 7.0
find_cuda_helper_libs(cusolver)
endif()
and also this piece of code replaced the old one :
set(CUDA_npp_LIBRARY "${CUDA_nppc_LIBRARY};${CUDA_nppial_LIBRARY};${CUDA_nppicc_LIBRARY};${CUDA_nppicom_LIBRARY};${CUDA_nppidei_LIBRARY};${CUDA_nppif_LIBRARY};${CUDA_nppig_LIBRARY};${CUDA_nppim_LIBRARY};${CUDA_nppist_LIBRARY};${CUDA_nppisu_LIBRARY};${CUDA_nppitc_LIBRARY};${CUDA_npps_LIBRARY}")
also this code replaced the old one :
unset(CUDA_nppial_LIBRARY CACHE)
unset(CUDA_nppicc_LIBRARY CACHE)
unset(CUDA_nppicom_LIBRARY CACHE)
unset(CUDA_nppidei_LIBRARY CACHE)
unset(CUDA_nppif_LIBRARY CACHE)
unset(CUDA_nppig_LIBRARY CACHE)
unset(CUDA_nppim_LIBRARY CACHE)
unset(CUDA_nppist_LIBRARY CACHE)
unset(CUDA_nppisu_LIBRARY CACHE)
unset(CUDA_nppitc_LIBRARY CACHE)
unset(CUDA_npps_LIBRARY CACHE)
unset(CUDA_nvcuvenc_LIBRARY CACHE)
unset(CUDA_nvcuvid_LIBRARY CACHE)
Regarding the other file,called "OpenCVDetectCUDA.cmake",I have also modified it exactly like explained there.
In the end,I have also added this line :
#include <cuda_fp16.h>
to this header file
/home/ziom/Scrivania/MonocularTotalCapture/opencv-2.4.13.6/modules/gpu/include/opencv2/gpu/device/common.hpp
but the error persists.
