generic-library/opencv
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

gpulegacy module fixes

Browse Source
This commit is contained in:
Vladislav Vinogradov
2013-04-18 11:30:54 +04:00
parent 6d735c11c6
commit 3be997cfa9
6 changed files with 43 additions and 60 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
modules/gpulegacy/src/cuda/NCVHaarObjectDetection.cu
View File

@@ -66,6 +66,7 @@
#ifdef HAVE_OPENCV_OBJDETECT
# include "opencv2/objdetect.hpp"
# include "opencv2/objdetect/objdetect_c.h"
#endif
#include "opencv2/gpulegacy/NCV.hpp"
@@ -2130,7 +2131,7 @@ static NCVStatus loadFromXML(const cv::String &filename,
haar.ClassifierSize.height = 0;
haar.bHasStumpsOnly = true;
haar.bNeedsTiltedII = false;
Ncv32u curMaxTreeDepth;
Ncv32u curMaxTreeDepth = 0;
std::vector<HaarClassifierNode128> h_TmpClassifierNotRootNodes;
haarStages.resize(0);

52
modules/gpulegacy/src/cuda/NCVPyramid.cu
View File

@@ -223,17 +223,25 @@ namespace cv { namespace gpu { namespace cudev
cudaSafeCall( cudaDeviceSynchronize() );
}
template void kernelDownsampleX2_gpu<uchar1>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
template void kernelDownsampleX2_gpu<uchar3>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
template void kernelDownsampleX2_gpu<uchar4>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
void downsampleX2(PtrStepSzb src, PtrStepSzb dst, int depth, int cn, cudaStream_t stream)
{
typedef void (*func_t)(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
template void kernelDownsampleX2_gpu<ushort1>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
template void kernelDownsampleX2_gpu<ushort3>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
template void kernelDownsampleX2_gpu<ushort4>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
static const func_t funcs[6][4] =
{
{kernelDownsampleX2_gpu<uchar1> , 0 /*kernelDownsampleX2_gpu<uchar2>*/ , kernelDownsampleX2_gpu<uchar3> , kernelDownsampleX2_gpu<uchar4> },
{0 /*kernelDownsampleX2_gpu<char1>*/ , 0 /*kernelDownsampleX2_gpu<char2>*/ , 0 /*kernelDownsampleX2_gpu<char3>*/ , 0 /*kernelDownsampleX2_gpu<char4>*/ },
{kernelDownsampleX2_gpu<ushort1> , 0 /*kernelDownsampleX2_gpu<ushort2>*/, kernelDownsampleX2_gpu<ushort3> , kernelDownsampleX2_gpu<ushort4> },
{0 /*kernelDownsampleX2_gpu<short1>*/ , 0 /*kernelDownsampleX2_gpu<short2>*/ , 0 /*kernelDownsampleX2_gpu<short3>*/, 0 /*kernelDownsampleX2_gpu<short4>*/},
{0 /*kernelDownsampleX2_gpu<int1>*/ , 0 /*kernelDownsampleX2_gpu<int2>*/ , 0 /*kernelDownsampleX2_gpu<int3>*/ , 0 /*kernelDownsampleX2_gpu<int4>*/ },
{kernelDownsampleX2_gpu<float1> , 0 /*kernelDownsampleX2_gpu<float2>*/ , kernelDownsampleX2_gpu<float3> , kernelDownsampleX2_gpu<float4> }
};
template void kernelDownsampleX2_gpu<float1>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
template void kernelDownsampleX2_gpu<float3>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
template void kernelDownsampleX2_gpu<float4>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
const func_t func = funcs[depth][cn - 1];
CV_Assert(func != 0);
func(src, dst, stream);
}
}
}}}
@@ -298,17 +306,25 @@ namespace cv { namespace gpu { namespace cudev
cudaSafeCall( cudaDeviceSynchronize() );
}
template void kernelInterpolateFrom1_gpu<uchar1>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
template void kernelInterpolateFrom1_gpu<uchar3>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
template void kernelInterpolateFrom1_gpu<uchar4>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
void interpolateFrom1(PtrStepSzb src, PtrStepSzb dst, int depth, int cn, cudaStream_t stream)
{
typedef void (*func_t)(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
template void kernelInterpolateFrom1_gpu<ushort1>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
template void kernelInterpolateFrom1_gpu<ushort3>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
template void kernelInterpolateFrom1_gpu<ushort4>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
static const func_t funcs[6][4] =
{
{kernelInterpolateFrom1_gpu<uchar1> , 0 /*kernelInterpolateFrom1_gpu<uchar2>*/ , kernelInterpolateFrom1_gpu<uchar3> , kernelInterpolateFrom1_gpu<uchar4> },
{0 /*kernelInterpolateFrom1_gpu<char1>*/ , 0 /*kernelInterpolateFrom1_gpu<char2>*/ , 0 /*kernelInterpolateFrom1_gpu<char3>*/ , 0 /*kernelInterpolateFrom1_gpu<char4>*/ },
{kernelInterpolateFrom1_gpu<ushort1> , 0 /*kernelInterpolateFrom1_gpu<ushort2>*/, kernelInterpolateFrom1_gpu<ushort3> , kernelInterpolateFrom1_gpu<ushort4> },
{0 /*kernelInterpolateFrom1_gpu<short1>*/, 0 /*kernelInterpolateFrom1_gpu<short2>*/ , 0 /*kernelInterpolateFrom1_gpu<short3>*/, 0 /*kernelInterpolateFrom1_gpu<short4>*/},
{0 /*kernelInterpolateFrom1_gpu<int1>*/ , 0 /*kernelInterpolateFrom1_gpu<int2>*/ , 0 /*kernelInterpolateFrom1_gpu<int3>*/ , 0 /*kernelInterpolateFrom1_gpu<int4>*/ },
{kernelInterpolateFrom1_gpu<float1> , 0 /*kernelInterpolateFrom1_gpu<float2>*/ , kernelInterpolateFrom1_gpu<float3> , kernelInterpolateFrom1_gpu<float4> }
};
template void kernelInterpolateFrom1_gpu<float1>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
template void kernelInterpolateFrom1_gpu<float3>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
template void kernelInterpolateFrom1_gpu<float4>(PtrStepSzb src, PtrStepSzb dst, cudaStream_t stream);
const func_t func = funcs[depth][cn - 1];
CV_Assert(func != 0);
func(src, dst, stream);
}
}
}}}