Geek Answers Handbook

OpenCV error at startup makes CentOS

I try to install OpenCV 2.4.2 on CentOS 5.8, but I get the following error

[ 34%] Built target opencv_test_highgui [ 34%] Building CXX object modules/features2d/CMakeFiles/opencv_features2d.dir/src/freak.cpp.o /home/jtrinidad/Downloads/OpenCV-2.4.2/modules/core/include/opencv2/core/core.hpp:1286: warning: 'class cv::_InputArray' has virtual functions but non-virtual destructor /home/jtrinidad/Downloads/OpenCV-2.4.2/modules/core/include/opencv2/core/core.hpp:1365: warning: 'class cv::_OutputArray' has virtual functions but non-virtual destructor /home/jtrinidad/Downloads/OpenCV-2.4.2/modules/features2d/src/freak.cpp: In member function 'virtual void cv::FREAK::computeImpl(const cv::Mat&, std::vector<cv::KeyPoint, std::allocator<cv::KeyPoint> >&, cv::Mat&) const': /home/jtrinidad/Downloads/OpenCV-2.4.2/modules/features2d/src/freak.cpp:367: error: shift must be an immediate make[2]: *** [modules/features2d/CMakeFiles/opencv_features2d.dir/src/freak.cpp.o] Error 1 make[1]: *** [modules/features2d/CMakeFiles/opencv_features2d.dir/all] Error 2 make: *** [all] Error 2

I ran the following command on cmake

  cmake -D CMAKE_BUILD_TYPE=RELEASE -D CMAKE_INSTALL_PREFIX=/usr/local -D BUILD_EXAMPLES=OFF WITH_CUDA=OFF ..

Thanks.

+4
source share
3 answers

The problem is resolved.

I applied the following patch to ~ / Downloads / OpenCV-2.4.2 / modules / features2d / src / freak.cpp

 diff -crB OpenCV-2.4.2.orig/modules/features2d/src/freak.cpp OpenCV-2.4.2/modules/features2d/src/freak.cpp *** OpenCV-2.4.2.orig/modules/features2d/src/freak.cpp 2012-07-25 09:54:50.000000000 -0400 --- OpenCV-2.4.2/modules/features2d/src/freak.cpp 2012-07-24 10:34:57.000000000 -0400 *************** *** 364,370 **** __m128i workReg = _mm_min_epu8(operand1, operand2); // emulated "not less than" for 8-bit UNSIGNED integers workReg = _mm_cmpeq_epi8(workReg, operand2); // emulated "not less than" for 8-bit UNSIGNED integers ! workReg = _mm_and_si128(_mm_srli_epi16(binMask, m), workReg); // merge the last 16 bits with the 128bits std::vector until full result128 = _mm_or_si128(result128, workReg); } (*ptr) = result128; --- 364,404 ---- __m128i workReg = _mm_min_epu8(operand1, operand2); // emulated "not less than" for 8-bit UNSIGNED integers workReg = _mm_cmpeq_epi8(workReg, operand2); // emulated "not less than" for 8-bit UNSIGNED integers ! // merge the last 16 bits with the 128bits std::vector until full ! __m128i shiftedMask; ! switch(m) { ! case 8: ! shiftedMask = _mm_srli_epi16(binMask, 8); ! break; ! case 7: ! shiftedMask = _mm_srli_epi16(binMask, 7); ! break; ! case 6: ! shiftedMask = _mm_srli_epi16(binMask, 6); ! break; ! case 5: ! shiftedMask = _mm_srli_epi16(binMask, 5); ! break; ! case 4: ! shiftedMask = _mm_srli_epi16(binMask, 4); ! break; ! case 3: ! shiftedMask = _mm_srli_epi16(binMask, 3); ! break; ! case 2: ! shiftedMask = _mm_srli_epi16(binMask, 2); ! break; ! case 1: ! shiftedMask = _mm_srli_epi16(binMask, 1); ! break; ! case 0: ! shiftedMask = _mm_srli_epi16(binMask, 0); ! break; ! default: ! throw "shifting less than 0"; ! } ! workReg = _mm_and_si128(shiftedMask, workReg); ! result128 = _mm_or_si128(result128, workReg); } (*ptr) = result128;
+4
source

After trying fixes, I resorted to compiling freak.cpp manually. My installation notes are available here: http://www.megalinux.net/compiling-opencv-2-4-on-rhelcentos-5/

0
source

More articles:


All Articles