Pablo Tello | c9564cb | 2019-09-13 10:20:25 +0100 | [diff] [blame] | 1 | /* |
Michele Di Giorgio | d9eaf61 | 2020-07-08 11:12:57 +0100 | [diff] [blame] | 2 | * Copyright (c) 2019-2020 Arm Limited. |
Pablo Tello | c9564cb | 2019-09-13 10:20:25 +0100 | [diff] [blame] | 3 | * |
| 4 | * SPDX-License-Identifier: MIT |
| 5 | * |
| 6 | * Permission is hereby granted, free of charge, to any person obtaining a copy |
| 7 | * of this software and associated documentation files (the "Software"), to |
| 8 | * deal in the Software without restriction, including without limitation the |
| 9 | * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or |
| 10 | * sell copies of the Software, and to permit persons to whom the Software is |
| 11 | * furnished to do so, subject to the following conditions: |
| 12 | * |
| 13 | * The above copyright notice and this permission notice shall be included in all |
| 14 | * copies or substantial portions of the Software. |
| 15 | * |
| 16 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
| 17 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
| 18 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE |
| 19 | * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
| 20 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
| 21 | * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
| 22 | * SOFTWARE. |
| 23 | */ |
| 24 | #include "arm_compute/runtime/NEON/NEScheduler.h" |
Pablo Tello | c9564cb | 2019-09-13 10:20:25 +0100 | [diff] [blame] | 25 | #include "arm_compute/runtime/NEON/functions/NEGenerateProposalsLayer.h" |
| 26 | #include "arm_compute/runtime/NEON/functions/NEPermute.h" |
| 27 | #include "arm_compute/runtime/NEON/functions/NESlice.h" |
Georgios Pinitas | 8c3c0e7 | 2020-12-03 20:11:53 +0000 | [diff] [blame] | 28 | #include "src/core/NEON/kernels/NEGenerateProposalsLayerKernel.h" |
Pablo Tello | c9564cb | 2019-09-13 10:20:25 +0100 | [diff] [blame] | 29 | #include "tests/Globals.h" |
| 30 | #include "tests/NEON/Accessor.h" |
| 31 | #include "tests/NEON/ArrayAccessor.h" |
Georgios Pinitas | 8c3c0e7 | 2020-12-03 20:11:53 +0000 | [diff] [blame] | 32 | #include "tests/NEON/Helper.h" |
Pablo Tello | c9564cb | 2019-09-13 10:20:25 +0100 | [diff] [blame] | 33 | #include "tests/framework/Macros.h" |
| 34 | #include "tests/framework/datasets/Datasets.h" |
| 35 | #include "tests/validation/Validation.h" |
| 36 | #include "tests/validation/fixtures/ComputeAllAnchorsFixture.h" |
| 37 | #include "utils/TypePrinter.h" |
| 38 | |
| 39 | namespace arm_compute |
| 40 | { |
| 41 | namespace test |
| 42 | { |
| 43 | namespace validation |
| 44 | { |
| 45 | namespace |
| 46 | { |
Georgios Pinitas | 8c3c0e7 | 2020-12-03 20:11:53 +0000 | [diff] [blame] | 47 | using NEComputeAllAnchors = NESynthetizeFunction<NEComputeAllAnchorsKernel>; |
| 48 | |
Pablo Tello | c9564cb | 2019-09-13 10:20:25 +0100 | [diff] [blame] | 49 | template <typename U, typename T> |
| 50 | inline void fill_tensor(U &&tensor, const std::vector<T> &v) |
| 51 | { |
| 52 | std::memcpy(tensor.data(), v.data(), sizeof(T) * v.size()); |
| 53 | } |
| 54 | |
| 55 | template <typename T> |
| 56 | inline void fill_tensor(Accessor &&tensor, const std::vector<T> &v) |
| 57 | { |
| 58 | if(tensor.data_layout() == DataLayout::NCHW) |
| 59 | { |
| 60 | std::memcpy(tensor.data(), v.data(), sizeof(T) * v.size()); |
| 61 | } |
| 62 | else |
| 63 | { |
| 64 | const int channels = tensor.shape()[0]; |
| 65 | const int width = tensor.shape()[1]; |
| 66 | const int height = tensor.shape()[2]; |
| 67 | for(int x = 0; x < width; ++x) |
| 68 | { |
| 69 | for(int y = 0; y < height; ++y) |
| 70 | { |
| 71 | for(int c = 0; c < channels; ++c) |
| 72 | { |
| 73 | *(reinterpret_cast<T *>(tensor(Coordinates(c, x, y)))) = *(reinterpret_cast<const T *>(v.data() + x + y * width + c * height * width)); |
| 74 | } |
| 75 | } |
| 76 | } |
| 77 | } |
| 78 | } |
| 79 | |
| 80 | const auto ComputeAllInfoDataset = framework::dataset::make("ComputeAllInfo", |
| 81 | { |
| 82 | ComputeAnchorsInfo(10U, 10U, 1. / 16.f), |
| 83 | ComputeAnchorsInfo(100U, 1U, 1. / 2.f), |
| 84 | ComputeAnchorsInfo(100U, 1U, 1. / 4.f), |
| 85 | ComputeAnchorsInfo(100U, 100U, 1. / 4.f), |
| 86 | |
| 87 | }); |
Michele Di Giorgio | 58c71ef | 2019-09-30 15:03:21 +0100 | [diff] [blame] | 88 | |
| 89 | constexpr AbsoluteTolerance<int16_t> tolerance_qsymm16(1); |
Pablo Tello | c9564cb | 2019-09-13 10:20:25 +0100 | [diff] [blame] | 90 | } // namespace |
| 91 | |
| 92 | TEST_SUITE(NEON) |
| 93 | TEST_SUITE(GenerateProposals) |
| 94 | |
| 95 | // *INDENT-OFF* |
| 96 | // clang-format off |
| 97 | DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip(zip(zip(zip(zip( |
| 98 | framework::dataset::make("scores", { TensorInfo(TensorShape(100U, 100U, 9U), 1, DataType::F32), |
| 99 | TensorInfo(TensorShape(100U, 100U, 9U), 1, DataType::F16), // Mismatching types |
| 100 | TensorInfo(TensorShape(100U, 100U, 9U), 1, DataType::F16), // Wrong deltas (number of transformation non multiple of 4) |
| 101 | TensorInfo(TensorShape(100U, 100U, 9U), 1, DataType::F16), // Wrong anchors (number of values per roi != 5) |
| 102 | TensorInfo(TensorShape(100U, 100U, 9U), 1, DataType::F16), // Output tensor num_valid_proposals not scalar |
| 103 | TensorInfo(TensorShape(100U, 100U, 9U), 1, DataType::F16)}), // num_valid_proposals not U32 |
| 104 | framework::dataset::make("deltas",{ TensorInfo(TensorShape(100U, 100U, 36U), 1, DataType::F32), |
| 105 | TensorInfo(TensorShape(100U, 100U, 36U), 1, DataType::F32), |
| 106 | TensorInfo(TensorShape(100U, 100U, 38U), 1, DataType::F32), |
| 107 | TensorInfo(TensorShape(100U, 100U, 38U), 1, DataType::F32), |
| 108 | TensorInfo(TensorShape(100U, 100U, 38U), 1, DataType::F32), |
| 109 | TensorInfo(TensorShape(100U, 100U, 38U), 1, DataType::F32)})), |
| 110 | framework::dataset::make("anchors", { TensorInfo(TensorShape(4U, 9U), 1, DataType::F32), |
| 111 | TensorInfo(TensorShape(4U, 9U), 1, DataType::F32), |
| 112 | TensorInfo(TensorShape(4U, 9U), 1, DataType::F32), |
| 113 | TensorInfo(TensorShape(5U, 9U), 1, DataType::F32), |
| 114 | TensorInfo(TensorShape(4U, 9U), 1, DataType::F32), |
| 115 | TensorInfo(TensorShape(4U, 9U), 1, DataType::F32)})), |
| 116 | framework::dataset::make("proposals", { TensorInfo(TensorShape(5U, 100U*100U*9U), 1, DataType::F32), |
| 117 | TensorInfo(TensorShape(5U, 100U*100U*9U), 1, DataType::F32), |
| 118 | TensorInfo(TensorShape(5U, 100U*100U*9U), 1, DataType::F32), |
| 119 | TensorInfo(TensorShape(5U, 100U*100U*9U), 1, DataType::F32), |
| 120 | TensorInfo(TensorShape(5U, 100U*100U*9U), 1, DataType::F32), |
| 121 | TensorInfo(TensorShape(5U, 100U*100U*9U), 1, DataType::F32)})), |
| 122 | framework::dataset::make("scores_out", { TensorInfo(TensorShape(100U*100U*9U), 1, DataType::F32), |
| 123 | TensorInfo(TensorShape(100U*100U*9U), 1, DataType::F32), |
| 124 | TensorInfo(TensorShape(100U*100U*9U), 1, DataType::F32), |
| 125 | TensorInfo(TensorShape(100U*100U*9U), 1, DataType::F32), |
| 126 | TensorInfo(TensorShape(100U*100U*9U), 1, DataType::F32), |
| 127 | TensorInfo(TensorShape(100U*100U*9U), 1, DataType::F32)})), |
| 128 | framework::dataset::make("num_valid_proposals", { TensorInfo(TensorShape(1U, 1U), 1, DataType::U32), |
| 129 | TensorInfo(TensorShape(1U, 1U), 1, DataType::U32), |
| 130 | TensorInfo(TensorShape(1U, 1U), 1, DataType::U32), |
| 131 | TensorInfo(TensorShape(1U, 1U), 1, DataType::U32), |
| 132 | TensorInfo(TensorShape(1U, 10U), 1, DataType::U32), |
| 133 | TensorInfo(TensorShape(1U, 1U), 1, DataType::F16)})), |
| 134 | framework::dataset::make("generate_proposals_info", { GenerateProposalsInfo(10.f, 10.f, 1.f), |
| 135 | GenerateProposalsInfo(10.f, 10.f, 1.f), |
| 136 | GenerateProposalsInfo(10.f, 10.f, 1.f), |
| 137 | GenerateProposalsInfo(10.f, 10.f, 1.f), |
| 138 | GenerateProposalsInfo(10.f, 10.f, 1.f), |
| 139 | GenerateProposalsInfo(10.f, 10.f, 1.f)})), |
| 140 | framework::dataset::make("Expected", { true, false, false, false, false, false })), |
| 141 | scores, deltas, anchors, proposals, scores_out, num_valid_proposals, generate_proposals_info, expected) |
| 142 | { |
| 143 | ARM_COMPUTE_EXPECT(bool(NEGenerateProposalsLayer::validate(&scores.clone()->set_is_resizable(true), |
| 144 | &deltas.clone()->set_is_resizable(true), |
| 145 | &anchors.clone()->set_is_resizable(true), |
| 146 | &proposals.clone()->set_is_resizable(true), |
| 147 | &scores_out.clone()->set_is_resizable(true), |
| 148 | &num_valid_proposals.clone()->set_is_resizable(true), |
| 149 | generate_proposals_info)) == expected, framework::LogLevel::ERRORS); |
| 150 | } |
| 151 | // clang-format on |
| 152 | // *INDENT-ON* |
| 153 | |
| 154 | template <typename T> |
| 155 | using NEComputeAllAnchorsFixture = ComputeAllAnchorsFixture<Tensor, Accessor, NEComputeAllAnchors, T>; |
| 156 | |
| 157 | TEST_SUITE(Float) |
| 158 | TEST_SUITE(FP32) |
| 159 | DATA_TEST_CASE(IntegrationTestCaseAllAnchors, framework::DatasetMode::ALL, framework::dataset::make("DataType", { DataType::F32 }), |
| 160 | data_type) |
| 161 | { |
| 162 | const int values_per_roi = 4; |
| 163 | const int num_anchors = 3; |
| 164 | const int feature_height = 4; |
| 165 | const int feature_width = 3; |
| 166 | |
| 167 | SimpleTensor<float> anchors_expected(TensorShape(values_per_roi, feature_width * feature_height * num_anchors), DataType::F32); |
| 168 | fill_tensor(anchors_expected, std::vector<float> { -26, -19, 87, 86, |
| 169 | -81, -27, 58, 63, |
| 170 | -44, -15, 55, 36, |
| 171 | -10, -19, 103, 86, |
| 172 | -65, -27, 74, 63, |
| 173 | -28, -15, 71, 36, |
| 174 | 6, -19, 119, 86, |
| 175 | -49, -27, 90, 63, |
| 176 | -12, -15, 87, 36, |
| 177 | -26, -3, 87, 102, |
| 178 | -81, -11, 58, 79, |
| 179 | -44, 1, 55, 52, |
| 180 | -10, -3, 103, 102, |
| 181 | -65, -11, 74, 79, |
| 182 | -28, 1, 71, 52, |
| 183 | 6, -3, 119, 102, |
| 184 | -49, -11, 90, 79, |
| 185 | -12, 1, 87, 52, |
| 186 | -26, 13, 87, 118, |
| 187 | -81, 5, 58, 95, |
| 188 | -44, 17, 55, 68, |
| 189 | -10, 13, 103, 118, |
| 190 | -65, 5, 74, 95, |
| 191 | -28, 17, 71, 68, |
| 192 | 6, 13, 119, 118, |
| 193 | -49, 5, 90, 95, |
| 194 | -12, 17, 87, 68, |
| 195 | -26, 29, 87, 134, |
| 196 | -81, 21, 58, 111, |
| 197 | -44, 33, 55, 84, |
| 198 | -10, 29, 103, 134, |
| 199 | -65, 21, 74, 111, |
| 200 | -28, 33, 71, 84, |
| 201 | 6, 29, 119, 134, |
| 202 | -49, 21, 90, 111, |
| 203 | -12, 33, 87, 84 |
| 204 | }); |
| 205 | |
| 206 | Tensor all_anchors; |
| 207 | Tensor anchors = create_tensor<Tensor>(TensorShape(4, num_anchors), data_type); |
| 208 | |
| 209 | // Create and configure function |
| 210 | NEComputeAllAnchors compute_anchors; |
| 211 | compute_anchors.configure(&anchors, &all_anchors, ComputeAnchorsInfo(feature_width, feature_height, 1. / 16.0)); |
| 212 | anchors.allocator()->allocate(); |
| 213 | all_anchors.allocator()->allocate(); |
| 214 | |
| 215 | fill_tensor(Accessor(anchors), std::vector<float> { -26, -19, 87, 86, |
| 216 | -81, -27, 58, 63, |
| 217 | -44, -15, 55, 36 |
| 218 | }); |
| 219 | // Compute function |
| 220 | compute_anchors.run(); |
| 221 | validate(Accessor(all_anchors), anchors_expected); |
| 222 | } |
| 223 | |
| 224 | DATA_TEST_CASE(IntegrationTestCaseGenerateProposals, framework::DatasetMode::ALL, combine(framework::dataset::make("DataType", { DataType::F32 }), |
| 225 | framework::dataset::make("DataLayout", { DataLayout::NCHW, DataLayout::NHWC })), |
| 226 | data_type, data_layout) |
| 227 | { |
| 228 | const int values_per_roi = 4; |
| 229 | const int num_anchors = 2; |
| 230 | const int feature_height = 4; |
| 231 | const int feature_width = 5; |
| 232 | |
| 233 | std::vector<float> scores_vector |
| 234 | { |
| 235 | 5.055894435664012e-04f, 1.270304909820112e-03f, 2.492271113912067e-03f, 5.951663827809190e-03f, |
| 236 | 7.846917156877404e-03f, 6.776275276294789e-03f, 6.761571012891965e-03f, 4.898292096237725e-03f, |
| 237 | 6.044472332578605e-04f, 3.203334118759474e-03f, 2.947527908919908e-03f, 6.313238560015770e-03f, |
| 238 | 7.931767757095738e-03f, 8.764345805102866e-03f, 7.325012199914913e-03f, 4.317069470446271e-03f, |
| 239 | 2.372537409795522e-03f, 1.589227460352735e-03f, 7.419477503600818e-03f, 3.157690354133824e-05f, |
| 240 | 1.125915135986472e-03f, 9.865363483872330e-03f, 2.429454743386769e-03f, 2.724460564167563e-03f, |
| 241 | 7.670409838207963e-03f, 5.558891552328172e-03f, 7.876904873099614e-03f, 6.824746047239291e-03f, |
| 242 | 7.023817548067892e-03f, 3.651314909238673e-04f, 6.720443709032501e-03f, 5.935615511606155e-03f, |
| 243 | 2.837349642759774e-03f, 1.787235113610299e-03f, 4.538568889918262e-03f, 3.391510678188818e-03f, |
| 244 | 7.328474239481874e-03f, 6.306967923936016e-03f, 8.102218904895860e-04f, 3.366646521610209e-03f |
| 245 | }; |
| 246 | |
| 247 | std::vector<float> bbx_vector |
| 248 | { |
| 249 | 5.066650471856862e-03, -7.638671742936328e-03, 2.549596503988635e-03, -8.316416756423296e-03, |
| 250 | -2.397471917924575e-04, 7.370595187754891e-03, -2.771880178185262e-03, 3.958364873973579e-03, |
| 251 | 4.493661094712284e-03, 2.016487051533088e-03, -5.893883038142033e-03, 7.570636080807809e-03, |
| 252 | -1.395511229386785e-03, 3.686686052704696e-03, -7.738166245767079e-03, -1.947306329828059e-03, |
| 253 | -9.299719716045681e-03, -3.476410493413708e-03, -2.390761190919604e-03, 4.359281254364210e-03, |
| 254 | -2.135251160164030e-04, 9.203299843371962e-03, 4.042322775006053e-03, -9.464271243910754e-03, |
| 255 | 2.566239543229305e-03, -9.691093900220627e-03, -4.019283034310979e-03, 8.145470429508792e-03, |
| 256 | 7.345087308315662e-04, 7.049642787384043e-03, -2.768492313674294e-03, 6.997160053405803e-03, |
| 257 | 6.675346697112969e-03, 2.353293365652274e-03, -3.612002585241749e-04, 1.592076522068768e-03, |
| 258 | -8.354188900818149e-04, -5.232515333564140e-04, 6.946683728847089e-03, -8.469757407935994e-03, |
| 259 | -8.985324496496555e-03, 4.885832859017961e-03, -7.662967577576512e-03, 7.284124004335807e-03, |
| 260 | -5.812167510299458e-03, -5.760336800482398e-03, 6.040416930336549e-03, 5.861508595443691e-03, |
| 261 | -5.509243096133549e-04, -2.006142470055888e-03, -7.205925340416066e-03, -1.117459082969758e-03, |
| 262 | 4.233247017623154e-03, 8.079257498201178e-03, 2.962639022639513e-03, 7.069474943472751e-03, |
| 263 | -8.562946284971293e-03, -8.228634642768271e-03, -6.116245322799971e-04, -7.213122000180859e-03, |
| 264 | 1.693094399433209e-03, -4.287504459132290e-03, 8.740365683925144e-03, 3.751788160720638e-03, |
| 265 | 7.006764222862830e-03, 9.676754678358187e-03, -6.458757235812945e-03, -4.486506575589758e-03, |
| 266 | -4.371087196816259e-03, 3.542166755953152e-03, -2.504808998699504e-03, 5.666601724512010e-03, |
| 267 | -3.691862724546129e-03, 3.689809719085287e-03, 9.079930264704458e-03, 6.365127787359476e-03, |
| 268 | 2.881681788246101e-06, 9.991866069315165e-03, -1.104757466496565e-03, -2.668455405633477e-03, |
| 269 | -1.225748887087659e-03, 6.530536159094015e-03, 3.629468917975644e-03, 1.374426066950348e-03, |
| 270 | -2.404098881570632e-03, -4.791365049441602e-03, -2.970654027009094e-03, 7.807553690294366e-03, |
| 271 | -1.198321129505323e-03, -3.574885336949881e-03, -5.380848303732298e-03, 9.705151282165116e-03, |
| 272 | -1.005217683242201e-03, 9.178094036278405e-03, -5.615977269541644e-03, 5.333533158509859e-03, |
| 273 | -2.817116206168516e-03, 6.672609782000503e-03, 6.575769501651313e-03, 8.987596634989362e-03, |
| 274 | -1.283530791296188e-03, 1.687717120057778e-03, 3.242391851439037e-03, -7.312060454341677e-03, |
| 275 | 4.735335326324270e-03, -6.832367028817463e-03, -5.414854835884652e-03, -9.352380213755996e-03, |
| 276 | -3.682662043703889e-03, -6.127508590419776e-04, -7.682256596819467e-03, 9.569532628790246e-03, |
| 277 | -1.572157284518933e-03, -6.023034366859191e-03, -5.110873282582924e-03, -8.697072236660256e-03, |
| 278 | -3.235150419663566e-03, -8.286320236471386e-03, -5.229472409112913e-03, 9.920785896115053e-03, |
| 279 | -2.478413362126123e-03, -9.261324796935007e-03, 1.718512310840434e-04, 3.015875488208480e-03, |
| 280 | -6.172932549255669e-03, -4.031715551985103e-03, -9.263878005853677e-03, -2.815310738453385e-03, |
| 281 | 7.075307462133643e-03, 1.404611747938669e-03, -1.518548732533266e-03, -9.293430941655778e-03, |
| 282 | 6.382186966633246e-03, 8.256835789169248e-03, 3.196907843506736e-03, 8.821615689753433e-03, |
| 283 | -7.661543424832439e-03, 1.636273081822326e-03, -8.792373335756125e-03, 2.958775812049877e-03, |
| 284 | -6.269300278071262e-03, 6.248285790856450e-03, -3.675414624536002e-03, -1.692616700318762e-03, |
| 285 | 4.126007647815893e-03, -9.155291689759584e-03, -8.432616039924004e-03, 4.899980636213323e-03, |
| 286 | 3.511535019681671e-03, -1.582745757177339e-03, -2.703657774917963e-03, 6.738168990840388e-03, |
| 287 | 4.300455303937919e-03, 9.618312854781494e-03, 2.762142918402472e-03, -6.590025003382154e-03, |
| 288 | -2.071168373801788e-03, 8.613893943683627e-03, 9.411190295341036e-03, -6.129018930548372e-03 |
| 289 | }; |
| 290 | |
| 291 | const std::vector<float> anchors_vector{ -26, -19, 87, 86, -81, -27, 58, 63 }; |
Giorgio Arena | 6e9d0e0 | 2020-01-03 15:02:04 +0000 | [diff] [blame] | 292 | SimpleTensor<float> proposals_expected(TensorShape(5, 9), DataType::F32); |
Pablo Tello | c9564cb | 2019-09-13 10:20:25 +0100 | [diff] [blame] | 293 | fill_tensor(proposals_expected, std::vector<float> |
| 294 | { |
| 295 | 0, 0, 0, 75.269, 64.4388, |
| 296 | 0, 21.9579, 13.0535, 119, 99, |
| 297 | 0, 38.303, 0, 119, 87.6447, |
| 298 | 0, 0, 0, 119, 64.619, |
| 299 | 0, 0, 20.7997, 74.0714, 99, |
| 300 | 0, 0, 0, 91.8963, 79.3724, |
| 301 | 0, 0, 4.42377, 58.1405, 95.1781, |
| 302 | 0, 0, 13.4405, 104.799, 99, |
| 303 | 0, 38.9066, 28.2434, 119, 99, |
| 304 | |
| 305 | }); |
| 306 | |
| 307 | SimpleTensor<float> scores_expected(TensorShape(9), DataType::F32); |
| 308 | fill_tensor(scores_expected, std::vector<float> |
| 309 | { |
| 310 | 0.00986536, |
| 311 | 0.00876435, |
| 312 | 0.00784692, |
| 313 | 0.00767041, |
| 314 | 0.00732847, |
| 315 | 0.00682475, |
| 316 | 0.00672044, |
| 317 | 0.00631324, |
| 318 | 3.15769e-05 |
| 319 | }); |
| 320 | |
| 321 | TensorShape scores_shape = TensorShape(feature_width, feature_height, num_anchors); |
| 322 | TensorShape deltas_shape = TensorShape(feature_width, feature_height, values_per_roi * num_anchors); |
| 323 | if(data_layout == DataLayout::NHWC) |
| 324 | { |
| 325 | permute(scores_shape, PermutationVector(2U, 0U, 1U)); |
| 326 | permute(deltas_shape, PermutationVector(2U, 0U, 1U)); |
| 327 | } |
| 328 | // Inputs |
| 329 | Tensor scores = create_tensor<Tensor>(scores_shape, data_type, 1, QuantizationInfo(), data_layout); |
| 330 | Tensor bbox_deltas = create_tensor<Tensor>(deltas_shape, data_type, 1, QuantizationInfo(), data_layout); |
| 331 | Tensor anchors = create_tensor<Tensor>(TensorShape(values_per_roi, num_anchors), data_type); |
| 332 | |
| 333 | // Outputs |
| 334 | Tensor proposals; |
| 335 | Tensor num_valid_proposals; |
| 336 | Tensor scores_out; |
| 337 | num_valid_proposals.allocator()->init(TensorInfo(TensorShape(1), 1, DataType::U32)); |
| 338 | |
| 339 | NEGenerateProposalsLayer generate_proposals; |
| 340 | generate_proposals.configure(&scores, &bbox_deltas, &anchors, &proposals, &scores_out, &num_valid_proposals, |
| 341 | GenerateProposalsInfo(120, 100, 0.166667f, 1 / 16.0, 6000, 300, 0.7f, 16.0f)); |
| 342 | |
| 343 | // Allocate memory for input/output tensors |
| 344 | scores.allocator()->allocate(); |
| 345 | bbox_deltas.allocator()->allocate(); |
| 346 | anchors.allocator()->allocate(); |
| 347 | proposals.allocator()->allocate(); |
| 348 | num_valid_proposals.allocator()->allocate(); |
| 349 | scores_out.allocator()->allocate(); |
| 350 | // Fill inputs |
| 351 | fill_tensor(Accessor(scores), scores_vector); |
| 352 | fill_tensor(Accessor(bbox_deltas), bbx_vector); |
| 353 | fill_tensor(Accessor(anchors), anchors_vector); |
| 354 | |
| 355 | // Run operator |
| 356 | generate_proposals.run(); |
| 357 | // Gather num_valid_proposals |
| 358 | const uint32_t N = *reinterpret_cast<uint32_t *>(num_valid_proposals.ptr_to_element(Coordinates(0, 0))); |
| 359 | |
| 360 | // Select the first N entries of the proposals |
| 361 | Tensor proposals_final; |
| 362 | NESlice select_proposals; |
| 363 | select_proposals.configure(&proposals, &proposals_final, Coordinates(0, 0), Coordinates(values_per_roi + 1, N)); |
| 364 | |
| 365 | proposals_final.allocator()->allocate(); |
| 366 | select_proposals.run(); |
| 367 | |
| 368 | // Select the first N entries of the proposals |
| 369 | Tensor scores_final; |
| 370 | NESlice select_scores; |
| 371 | select_scores.configure(&scores_out, &scores_final, Coordinates(0), Coordinates(N)); |
| 372 | scores_final.allocator()->allocate(); |
| 373 | select_scores.run(); |
| 374 | |
| 375 | const RelativeTolerance<float> tolerance_f32(1e-5f); |
| 376 | // Validate the output |
| 377 | validate(Accessor(proposals_final), proposals_expected, tolerance_f32); |
| 378 | validate(Accessor(scores_final), scores_expected, tolerance_f32); |
| 379 | } |
| 380 | |
| 381 | FIXTURE_DATA_TEST_CASE(ComputeAllAnchors, NEComputeAllAnchorsFixture<float>, framework::DatasetMode::ALL, |
| 382 | combine(combine(framework::dataset::make("NumAnchors", { 2, 4, 8 }), ComputeAllInfoDataset), framework::dataset::make("DataType", { DataType::F32 }))) |
| 383 | { |
| 384 | // Validate output |
| 385 | validate(Accessor(_target), _reference); |
| 386 | } |
| 387 | TEST_SUITE_END() // FP32 |
| 388 | #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC |
| 389 | TEST_SUITE(FP16) |
| 390 | FIXTURE_DATA_TEST_CASE(ComputeAllAnchors, NEComputeAllAnchorsFixture<half>, framework::DatasetMode::ALL, |
| 391 | combine(combine(framework::dataset::make("NumAnchors", { 2, 4, 8 }), ComputeAllInfoDataset), framework::dataset::make("DataType", { DataType::F16 }))) |
| 392 | { |
| 393 | // Validate output |
| 394 | validate(Accessor(_target), _reference); |
| 395 | } |
| 396 | TEST_SUITE_END() // FP16 |
| 397 | #endif // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC |
| 398 | |
| 399 | TEST_SUITE_END() // Float |
| 400 | |
Michele Di Giorgio | 58c71ef | 2019-09-30 15:03:21 +0100 | [diff] [blame] | 401 | template <typename T> |
| 402 | using NEComputeAllAnchorsQuantizedFixture = ComputeAllAnchorsQuantizedFixture<Tensor, Accessor, NEComputeAllAnchors, T>; |
| 403 | |
| 404 | TEST_SUITE(Quantized) |
| 405 | TEST_SUITE(QASYMM8) |
| 406 | FIXTURE_DATA_TEST_CASE(ComputeAllAnchors, NEComputeAllAnchorsQuantizedFixture<int16_t>, framework::DatasetMode::ALL, |
| 407 | combine(combine(combine(framework::dataset::make("NumAnchors", { 2, 4, 8 }), ComputeAllInfoDataset), |
| 408 | framework::dataset::make("DataType", { DataType::QSYMM16 })), |
| 409 | framework::dataset::make("QuantInfo", { QuantizationInfo(0.125f, 0) }))) |
| 410 | { |
| 411 | // Validate output |
| 412 | validate(Accessor(_target), _reference, tolerance_qsymm16); |
| 413 | } |
| 414 | TEST_SUITE_END() // QASYMM8 |
| 415 | TEST_SUITE_END() // Quantized |
| 416 | |
Pablo Tello | c9564cb | 2019-09-13 10:20:25 +0100 | [diff] [blame] | 417 | TEST_SUITE_END() // GenerateProposals |
| 418 | TEST_SUITE_END() // NEON |
Pablo Tello | c9564cb | 2019-09-13 10:20:25 +0100 | [diff] [blame] | 419 | } // namespace validation |
| 420 | } // namespace test |
| 421 | } // namespace arm_compute |