1 /*M/////////////////////////////////////////////////////////////////////////////////////// 2 // 3 // IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING. 4 // 5 // By downloading, copying, installing or using the software you agree to this license. 6 // If you do not agree to this license, do not download, install, 7 // copy or use the software. 8 // 9 // 10 // License Agreement 11 // For Open Source Computer Vision Library 12 // 13 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved. 14 // Copyright (C) 2009, Willow Garage Inc., all rights reserved. 15 // Third party copyrights are property of their respective owners. 16 // 17 // Redistribution and use in source and binary forms, with or without modification, 18 // are permitted provided that the following conditions are met: 19 // 20 // * Redistribution's of source code must retain the above copyright notice, 21 // this list of conditions and the following disclaimer. 22 // 23 // * Redistribution's in binary form must reproduce the above copyright notice, 24 // this list of conditions and the following disclaimer in the documentation 25 // and/or other materials provided with the distribution. 26 // 27 // * The name of the copyright holders may not be used to endorse or promote products 28 // derived from this software without specific prior written permission. 29 // 30 // This software is provided by the copyright holders and contributors "as is" and 31 // any express or implied warranties, including, but not limited to, the implied 32 // warranties of merchantability and fitness for a particular purpose are disclaimed. 33 // In no event shall the Intel Corporation or contributors be liable for any direct, 34 // indirect, incidental, special, exemplary, or consequential damages 35 // (including, but not limited to, procurement of substitute goods or services; 36 // loss of use, data, or profits; or business interruption) however caused 37 // and on any theory of liability, whether in contract, strict liability, 38 // or tort (including negligence or otherwise) arising in any way out of 39 // the use of this software, even if advised of the possibility of such damage. 40 // 41 //M*/ 42 43 #if !defined CUDA_DISABLER 44 45 #include "opencv2/core/cuda/common.hpp" 46 #include "opencv2/core/cuda/vec_traits.hpp" 47 #include "opencv2/core/cuda/limits.hpp" 48 49 namespace cv { namespace cuda { namespace device { 50 namespace gmg 51 { 52 __constant__ int c_width; 53 __constant__ int c_height; 54 __constant__ float c_minVal; 55 __constant__ float c_maxVal; 56 __constant__ int c_quantizationLevels; 57 __constant__ float c_backgroundPrior; 58 __constant__ float c_decisionThreshold; 59 __constant__ int c_maxFeatures; 60 __constant__ int c_numInitializationFrames; 61 loadConstants(int width,int height,float minVal,float maxVal,int quantizationLevels,float backgroundPrior,float decisionThreshold,int maxFeatures,int numInitializationFrames)62 void loadConstants(int width, int height, float minVal, float maxVal, int quantizationLevels, float backgroundPrior, 63 float decisionThreshold, int maxFeatures, int numInitializationFrames) 64 { 65 cudaSafeCall( cudaMemcpyToSymbol(c_width, &width, sizeof(width)) ); 66 cudaSafeCall( cudaMemcpyToSymbol(c_height, &height, sizeof(height)) ); 67 cudaSafeCall( cudaMemcpyToSymbol(c_minVal, &minVal, sizeof(minVal)) ); 68 cudaSafeCall( cudaMemcpyToSymbol(c_maxVal, &maxVal, sizeof(maxVal)) ); 69 cudaSafeCall( cudaMemcpyToSymbol(c_quantizationLevels, &quantizationLevels, sizeof(quantizationLevels)) ); 70 cudaSafeCall( cudaMemcpyToSymbol(c_backgroundPrior, &backgroundPrior, sizeof(backgroundPrior)) ); 71 cudaSafeCall( cudaMemcpyToSymbol(c_decisionThreshold, &decisionThreshold, sizeof(decisionThreshold)) ); 72 cudaSafeCall( cudaMemcpyToSymbol(c_maxFeatures, &maxFeatures, sizeof(maxFeatures)) ); 73 cudaSafeCall( cudaMemcpyToSymbol(c_numInitializationFrames, &numInitializationFrames, sizeof(numInitializationFrames)) ); 74 } 75 findFeature(const int color,const PtrStepi & colors,const PtrStepf & weights,const int x,const int y,const int nfeatures)76 __device__ float findFeature(const int color, const PtrStepi& colors, const PtrStepf& weights, const int x, const int y, const int nfeatures) 77 { 78 for (int i = 0, fy = y; i < nfeatures; ++i, fy += c_height) 79 { 80 if (color == colors(fy, x)) 81 return weights(fy, x); 82 } 83 84 // not in histogram, so return 0. 85 return 0.0f; 86 } 87 normalizeHistogram(PtrStepf weights,const int x,const int y,const int nfeatures)88 __device__ void normalizeHistogram(PtrStepf weights, const int x, const int y, const int nfeatures) 89 { 90 float total = 0.0f; 91 for (int i = 0, fy = y; i < nfeatures; ++i, fy += c_height) 92 total += weights(fy, x); 93 94 if (total != 0.0f) 95 { 96 for (int i = 0, fy = y; i < nfeatures; ++i, fy += c_height) 97 weights(fy, x) /= total; 98 } 99 } 100 insertFeature(const int color,const float weight,PtrStepi colors,PtrStepf weights,const int x,const int y,int & nfeatures)101 __device__ bool insertFeature(const int color, const float weight, PtrStepi colors, PtrStepf weights, const int x, const int y, int& nfeatures) 102 { 103 for (int i = 0, fy = y; i < nfeatures; ++i, fy += c_height) 104 { 105 if (color == colors(fy, x)) 106 { 107 // feature in histogram 108 109 weights(fy, x) += weight; 110 111 return false; 112 } 113 } 114 115 if (nfeatures == c_maxFeatures) 116 { 117 // discard oldest feature 118 119 int idx = -1; 120 float minVal = numeric_limits<float>::max(); 121 for (int i = 0, fy = y; i < nfeatures; ++i, fy += c_height) 122 { 123 const float w = weights(fy, x); 124 if (w < minVal) 125 { 126 minVal = w; 127 idx = fy; 128 } 129 } 130 131 colors(idx, x) = color; 132 weights(idx, x) = weight; 133 134 return false; 135 } 136 137 colors(nfeatures * c_height + y, x) = color; 138 weights(nfeatures * c_height + y, x) = weight; 139 140 ++nfeatures; 141 142 return true; 143 } 144 145 namespace detail 146 { 147 template <int cn> struct Quantization 148 { 149 template <typename T> applycv::cuda::device::gmg::detail::Quantization150 __device__ static int apply(const T& val) 151 { 152 int res = 0; 153 res |= static_cast<int>((val.x - c_minVal) * c_quantizationLevels / (c_maxVal - c_minVal)); 154 res |= static_cast<int>((val.y - c_minVal) * c_quantizationLevels / (c_maxVal - c_minVal)) << 8; 155 res |= static_cast<int>((val.z - c_minVal) * c_quantizationLevels / (c_maxVal - c_minVal)) << 16; 156 return res; 157 } 158 }; 159 160 template <> struct Quantization<1> 161 { 162 template <typename T> applycv::cuda::device::gmg::detail::Quantization163 __device__ static int apply(T val) 164 { 165 return static_cast<int>((val - c_minVal) * c_quantizationLevels / (c_maxVal - c_minVal)); 166 } 167 }; 168 } 169 170 template <typename T> struct Quantization : detail::Quantization<VecTraits<T>::cn> {}; 171 172 template <typename SrcT> update(const PtrStep<SrcT> frame,PtrStepb fgmask,PtrStepi colors_,PtrStepf weights_,PtrStepi nfeatures_,const int frameNum,const float learningRate,const bool updateBackgroundModel)173 __global__ void update(const PtrStep<SrcT> frame, PtrStepb fgmask, PtrStepi colors_, PtrStepf weights_, PtrStepi nfeatures_, 174 const int frameNum, const float learningRate, const bool updateBackgroundModel) 175 { 176 const int x = blockIdx.x * blockDim.x + threadIdx.x; 177 const int y = blockIdx.y * blockDim.y + threadIdx.y; 178 179 if (x >= c_width || y >= c_height) 180 return; 181 182 const SrcT pix = frame(y, x); 183 const int newFeatureColor = Quantization<SrcT>::apply(pix); 184 185 int nfeatures = nfeatures_(y, x); 186 187 if (frameNum >= c_numInitializationFrames) 188 { 189 // typical operation 190 191 const float weight = findFeature(newFeatureColor, colors_, weights_, x, y, nfeatures); 192 193 // see Godbehere, Matsukawa, Goldberg (2012) for reasoning behind this implementation of Bayes rule 194 const float posterior = (weight * c_backgroundPrior) / (weight * c_backgroundPrior + (1.0f - weight) * (1.0f - c_backgroundPrior)); 195 196 const bool isForeground = ((1.0f - posterior) > c_decisionThreshold); 197 fgmask(y, x) = (uchar)(-isForeground); 198 199 // update histogram. 200 201 if (updateBackgroundModel) 202 { 203 for (int i = 0, fy = y; i < nfeatures; ++i, fy += c_height) 204 weights_(fy, x) *= 1.0f - learningRate; 205 206 bool inserted = insertFeature(newFeatureColor, learningRate, colors_, weights_, x, y, nfeatures); 207 208 if (inserted) 209 { 210 normalizeHistogram(weights_, x, y, nfeatures); 211 nfeatures_(y, x) = nfeatures; 212 } 213 } 214 } 215 else if (updateBackgroundModel) 216 { 217 // training-mode update 218 219 insertFeature(newFeatureColor, 1.0f, colors_, weights_, x, y, nfeatures); 220 221 if (frameNum == c_numInitializationFrames - 1) 222 normalizeHistogram(weights_, x, y, nfeatures); 223 } 224 } 225 226 template <typename SrcT> update_gpu(PtrStepSzb frame,PtrStepb fgmask,PtrStepSzi colors,PtrStepf weights,PtrStepi nfeatures,int frameNum,float learningRate,bool updateBackgroundModel,cudaStream_t stream)227 void update_gpu(PtrStepSzb frame, PtrStepb fgmask, PtrStepSzi colors, PtrStepf weights, PtrStepi nfeatures, 228 int frameNum, float learningRate, bool updateBackgroundModel, cudaStream_t stream) 229 { 230 const dim3 block(32, 8); 231 const dim3 grid(divUp(frame.cols, block.x), divUp(frame.rows, block.y)); 232 233 cudaSafeCall( cudaFuncSetCacheConfig(update<SrcT>, cudaFuncCachePreferL1) ); 234 235 update<SrcT><<<grid, block, 0, stream>>>((PtrStepSz<SrcT>) frame, fgmask, colors, weights, nfeatures, frameNum, learningRate, updateBackgroundModel); 236 237 cudaSafeCall( cudaGetLastError() ); 238 239 if (stream == 0) 240 cudaSafeCall( cudaDeviceSynchronize() ); 241 } 242 243 template void update_gpu<uchar >(PtrStepSzb frame, PtrStepb fgmask, PtrStepSzi colors, PtrStepf weights, PtrStepi nfeatures, int frameNum, float learningRate, bool updateBackgroundModel, cudaStream_t stream); 244 template void update_gpu<uchar3 >(PtrStepSzb frame, PtrStepb fgmask, PtrStepSzi colors, PtrStepf weights, PtrStepi nfeatures, int frameNum, float learningRate, bool updateBackgroundModel, cudaStream_t stream); 245 template void update_gpu<uchar4 >(PtrStepSzb frame, PtrStepb fgmask, PtrStepSzi colors, PtrStepf weights, PtrStepi nfeatures, int frameNum, float learningRate, bool updateBackgroundModel, cudaStream_t stream); 246 247 template void update_gpu<ushort >(PtrStepSzb frame, PtrStepb fgmask, PtrStepSzi colors, PtrStepf weights, PtrStepi nfeatures, int frameNum, float learningRate, bool updateBackgroundModel, cudaStream_t stream); 248 template void update_gpu<ushort3>(PtrStepSzb frame, PtrStepb fgmask, PtrStepSzi colors, PtrStepf weights, PtrStepi nfeatures, int frameNum, float learningRate, bool updateBackgroundModel, cudaStream_t stream); 249 template void update_gpu<ushort4>(PtrStepSzb frame, PtrStepb fgmask, PtrStepSzi colors, PtrStepf weights, PtrStepi nfeatures, int frameNum, float learningRate, bool updateBackgroundModel, cudaStream_t stream); 250 251 template void update_gpu<float >(PtrStepSzb frame, PtrStepb fgmask, PtrStepSzi colors, PtrStepf weights, PtrStepi nfeatures, int frameNum, float learningRate, bool updateBackgroundModel, cudaStream_t stream); 252 template void update_gpu<float3 >(PtrStepSzb frame, PtrStepb fgmask, PtrStepSzi colors, PtrStepf weights, PtrStepi nfeatures, int frameNum, float learningRate, bool updateBackgroundModel, cudaStream_t stream); 253 template void update_gpu<float4 >(PtrStepSzb frame, PtrStepb fgmask, PtrStepSzi colors, PtrStepf weights, PtrStepi nfeatures, int frameNum, float learningRate, bool updateBackgroundModel, cudaStream_t stream); 254 } 255 }}} 256 257 258 #endif /* CUDA_DISABLER */ 259