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 #include "perf_precomp.hpp" 44 45 using namespace std; 46 using namespace testing; 47 using namespace perf; 48 49 ////////////////////////////////////////////////////////////////////// 50 // FAST 51 52 DEF_PARAM_TEST(Image_Threshold_NonMaxSuppression, string, int, bool); 53 54 PERF_TEST_P(Image_Threshold_NonMaxSuppression, FAST, 55 Combine(Values<string>("gpu/perf/aloe.png"), 56 Values(20), 57 Bool())) 58 { 59 const cv::Mat img = readImage(GET_PARAM(0), cv::IMREAD_GRAYSCALE); 60 ASSERT_FALSE(img.empty()); 61 62 const int threshold = GET_PARAM(1); 63 const bool nonMaxSuppersion = GET_PARAM(2); 64 65 if (PERF_RUN_CUDA()) 66 { 67 cv::Ptr<cv::cuda::FastFeatureDetector> d_fast = 68 cv::cuda::FastFeatureDetector::create(threshold, nonMaxSuppersion, 69 cv::FastFeatureDetector::TYPE_9_16, 70 0.5 * img.size().area()); 71 72 const cv::cuda::GpuMat d_img(img); 73 cv::cuda::GpuMat d_keypoints; 74 75 TEST_CYCLE() d_fast->detectAsync(d_img, d_keypoints); 76 77 std::vector<cv::KeyPoint> gpu_keypoints; 78 d_fast->convert(d_keypoints, gpu_keypoints); 79 80 sortKeyPoints(gpu_keypoints); 81 82 SANITY_CHECK_KEYPOINTS(gpu_keypoints); 83 } 84 else 85 { 86 std::vector<cv::KeyPoint> cpu_keypoints; 87 88 TEST_CYCLE() cv::FAST(img, cpu_keypoints, threshold, nonMaxSuppersion); 89 90 SANITY_CHECK_KEYPOINTS(cpu_keypoints); 91 } 92 } 93 94 ////////////////////////////////////////////////////////////////////// 95 // ORB 96 97 DEF_PARAM_TEST(Image_NFeatures, string, int); 98 99 PERF_TEST_P(Image_NFeatures, ORB, 100 Combine(Values<string>("gpu/perf/aloe.png"), 101 Values(4000))) 102 { 103 declare.time(300.0); 104 105 const cv::Mat img = readImage(GET_PARAM(0), cv::IMREAD_GRAYSCALE); 106 ASSERT_FALSE(img.empty()); 107 108 const int nFeatures = GET_PARAM(1); 109 110 if (PERF_RUN_CUDA()) 111 { 112 cv::Ptr<cv::cuda::ORB> d_orb = cv::cuda::ORB::create(nFeatures); 113 114 const cv::cuda::GpuMat d_img(img); 115 cv::cuda::GpuMat d_keypoints, d_descriptors; 116 117 TEST_CYCLE() d_orb->detectAndComputeAsync(d_img, cv::noArray(), d_keypoints, d_descriptors); 118 119 std::vector<cv::KeyPoint> gpu_keypoints; 120 d_orb->convert(d_keypoints, gpu_keypoints); 121 122 cv::Mat gpu_descriptors(d_descriptors); 123 124 gpu_keypoints.resize(10); 125 gpu_descriptors = gpu_descriptors.rowRange(0, 10); 126 127 sortKeyPoints(gpu_keypoints, gpu_descriptors); 128 129 SANITY_CHECK_KEYPOINTS(gpu_keypoints, 1e-4); 130 SANITY_CHECK(gpu_descriptors); 131 } 132 else 133 { 134 cv::Ptr<cv::ORB> orb = cv::ORB::create(nFeatures); 135 136 std::vector<cv::KeyPoint> cpu_keypoints; 137 cv::Mat cpu_descriptors; 138 139 TEST_CYCLE() orb->detectAndCompute(img, cv::noArray(), cpu_keypoints, cpu_descriptors); 140 141 SANITY_CHECK_KEYPOINTS(cpu_keypoints); 142 SANITY_CHECK(cpu_descriptors); 143 } 144 } 145 146 ////////////////////////////////////////////////////////////////////// 147 // BFMatch 148 149 DEF_PARAM_TEST(DescSize_Norm, int, NormType); 150 151 PERF_TEST_P(DescSize_Norm, BFMatch, 152 Combine(Values(64, 128, 256), 153 Values(NormType(cv::NORM_L1), NormType(cv::NORM_L2), NormType(cv::NORM_HAMMING)))) 154 { 155 declare.time(20.0); 156 157 const int desc_size = GET_PARAM(0); 158 const int normType = GET_PARAM(1); 159 160 const int type = normType == cv::NORM_HAMMING ? CV_8U : CV_32F; 161 162 cv::Mat query(3000, desc_size, type); 163 declare.in(query, WARMUP_RNG); 164 165 cv::Mat train(3000, desc_size, type); 166 declare.in(train, WARMUP_RNG); 167 168 if (PERF_RUN_CUDA()) 169 { 170 cv::Ptr<cv::cuda::DescriptorMatcher> d_matcher = cv::cuda::DescriptorMatcher::createBFMatcher(normType); 171 172 const cv::cuda::GpuMat d_query(query); 173 const cv::cuda::GpuMat d_train(train); 174 cv::cuda::GpuMat d_matches; 175 176 TEST_CYCLE() d_matcher->matchAsync(d_query, d_train, d_matches); 177 178 std::vector<cv::DMatch> gpu_matches; 179 d_matcher->matchConvert(d_matches, gpu_matches); 180 181 SANITY_CHECK_MATCHES(gpu_matches); 182 } 183 else 184 { 185 cv::BFMatcher matcher(normType); 186 187 std::vector<cv::DMatch> cpu_matches; 188 189 TEST_CYCLE() matcher.match(query, train, cpu_matches); 190 191 SANITY_CHECK_MATCHES(cpu_matches); 192 } 193 } 194 195 ////////////////////////////////////////////////////////////////////// 196 // BFKnnMatch 197 198 static void toOneRowMatches(const std::vector< std::vector<cv::DMatch> >& src, std::vector<cv::DMatch>& dst) 199 { 200 dst.clear(); 201 for (size_t i = 0; i < src.size(); ++i) 202 for (size_t j = 0; j < src[i].size(); ++j) 203 dst.push_back(src[i][j]); 204 } 205 206 DEF_PARAM_TEST(DescSize_K_Norm, int, int, NormType); 207 208 PERF_TEST_P(DescSize_K_Norm, BFKnnMatch, 209 Combine(Values(64, 128, 256), 210 Values(2, 3), 211 Values(NormType(cv::NORM_L1), NormType(cv::NORM_L2)))) 212 { 213 declare.time(30.0); 214 215 const int desc_size = GET_PARAM(0); 216 const int k = GET_PARAM(1); 217 const int normType = GET_PARAM(2); 218 219 const int type = normType == cv::NORM_HAMMING ? CV_8U : CV_32F; 220 221 cv::Mat query(3000, desc_size, type); 222 declare.in(query, WARMUP_RNG); 223 224 cv::Mat train(3000, desc_size, type); 225 declare.in(train, WARMUP_RNG); 226 227 if (PERF_RUN_CUDA()) 228 { 229 cv::Ptr<cv::cuda::DescriptorMatcher> d_matcher = cv::cuda::DescriptorMatcher::createBFMatcher(normType); 230 231 const cv::cuda::GpuMat d_query(query); 232 const cv::cuda::GpuMat d_train(train); 233 cv::cuda::GpuMat d_matches; 234 235 TEST_CYCLE() d_matcher->knnMatchAsync(d_query, d_train, d_matches, k); 236 237 std::vector< std::vector<cv::DMatch> > matchesTbl; 238 d_matcher->knnMatchConvert(d_matches, matchesTbl); 239 240 std::vector<cv::DMatch> gpu_matches; 241 toOneRowMatches(matchesTbl, gpu_matches); 242 243 SANITY_CHECK_MATCHES(gpu_matches); 244 } 245 else 246 { 247 cv::BFMatcher matcher(normType); 248 249 std::vector< std::vector<cv::DMatch> > matchesTbl; 250 251 TEST_CYCLE() matcher.knnMatch(query, train, matchesTbl, k); 252 253 std::vector<cv::DMatch> cpu_matches; 254 toOneRowMatches(matchesTbl, cpu_matches); 255 256 SANITY_CHECK_MATCHES(cpu_matches); 257 } 258 } 259 260 ////////////////////////////////////////////////////////////////////// 261 // BFRadiusMatch 262 263 PERF_TEST_P(DescSize_Norm, BFRadiusMatch, 264 Combine(Values(64, 128, 256), 265 Values(NormType(cv::NORM_L1), NormType(cv::NORM_L2)))) 266 { 267 declare.time(30.0); 268 269 const int desc_size = GET_PARAM(0); 270 const int normType = GET_PARAM(1); 271 272 const int type = normType == cv::NORM_HAMMING ? CV_8U : CV_32F; 273 const float maxDistance = 10000; 274 275 cv::Mat query(3000, desc_size, type); 276 declare.in(query, WARMUP_RNG); 277 278 cv::Mat train(3000, desc_size, type); 279 declare.in(train, WARMUP_RNG); 280 281 if (PERF_RUN_CUDA()) 282 { 283 cv::Ptr<cv::cuda::DescriptorMatcher> d_matcher = cv::cuda::DescriptorMatcher::createBFMatcher(normType); 284 285 const cv::cuda::GpuMat d_query(query); 286 const cv::cuda::GpuMat d_train(train); 287 cv::cuda::GpuMat d_matches; 288 289 TEST_CYCLE() d_matcher->radiusMatchAsync(d_query, d_train, d_matches, maxDistance); 290 291 std::vector< std::vector<cv::DMatch> > matchesTbl; 292 d_matcher->radiusMatchConvert(d_matches, matchesTbl); 293 294 std::vector<cv::DMatch> gpu_matches; 295 toOneRowMatches(matchesTbl, gpu_matches); 296 297 SANITY_CHECK_MATCHES(gpu_matches); 298 } 299 else 300 { 301 cv::BFMatcher matcher(normType); 302 303 std::vector< std::vector<cv::DMatch> > matchesTbl; 304 305 TEST_CYCLE() matcher.radiusMatch(query, train, matchesTbl, maxDistance); 306 307 std::vector<cv::DMatch> cpu_matches; 308 toOneRowMatches(matchesTbl, cpu_matches); 309 310 SANITY_CHECK_MATCHES(cpu_matches); 311 } 312 } 313
