1 /* 2 * Copyright (c) 2021 Arm Limited. 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 #ifndef ARM_COMPUTE_TEST_ROIPOOLINGLAYER_FIXTURE 25 #define ARM_COMPUTE_TEST_ROIPOOLINGLAYER_FIXTURE 26 27 #include "arm_compute/core/TensorShape.h" 28 #include "arm_compute/core/Types.h" 29 #include "arm_compute/core/utils/misc/ShapeCalculator.h" 30 #include "tests/AssetsLibrary.h" 31 #include "tests/Globals.h" 32 #include "tests/IAccessor.h" 33 #include "tests/framework/Asserts.h" 34 #include "tests/framework/Fixture.h" 35 #include "tests/validation/Helpers.h" 36 #include "tests/validation/reference/ROIPoolingLayer.h" 37 38 namespace arm_compute 39 { 40 namespace test 41 { 42 namespace validation 43 { 44 template <typename TensorType, typename AccessorType, typename FunctionType, typename T> 45 class ROIPoolingLayerGenericFixture : public framework::Fixture 46 { 47 public: 48 template <typename...> setup(TensorShape input_shape,const ROIPoolingLayerInfo pool_info,TensorShape rois_shape,DataType data_type,DataLayout data_layout,QuantizationInfo qinfo,QuantizationInfo output_qinfo)49 void setup(TensorShape input_shape, const ROIPoolingLayerInfo pool_info, TensorShape rois_shape, DataType data_type, DataLayout data_layout, QuantizationInfo qinfo, QuantizationInfo output_qinfo) 50 { 51 _target = compute_target(input_shape, data_type, data_layout, pool_info, rois_shape, qinfo, output_qinfo); 52 _reference = compute_reference(input_shape, data_type, pool_info, rois_shape, qinfo, output_qinfo); 53 } 54 55 protected: 56 template <typename U> fill(U && tensor)57 void fill(U &&tensor) 58 { 59 library->fill_tensor_uniform(tensor, 0); 60 } 61 62 template <typename U> 63 void generate_rois(U &&rois, const TensorShape &shape, const ROIPoolingLayerInfo &pool_info, TensorShape rois_shape, DataLayout data_layout = DataLayout::NCHW) 64 { 65 const size_t values_per_roi = rois_shape.x(); 66 const size_t num_rois = rois_shape.y(); 67 68 std::mt19937 gen(library->seed()); 69 uint16_t *rois_ptr = static_cast<uint16_t *>(rois.data()); 70 71 const float pool_width = pool_info.pooled_width(); 72 const float pool_height = pool_info.pooled_height(); 73 const float roi_scale = pool_info.spatial_scale(); 74 75 // Calculate distribution bounds 76 const auto scaled_width = static_cast<float>((shape[get_data_layout_dimension_index(data_layout, DataLayoutDimension::WIDTH)] / roi_scale) / pool_width); 77 const auto scaled_height = static_cast<float>((shape[get_data_layout_dimension_index(data_layout, DataLayoutDimension::HEIGHT)] / roi_scale) / pool_height); 78 const auto min_width = static_cast<float>(pool_width / roi_scale); 79 const auto min_height = static_cast<float>(pool_height / roi_scale); 80 81 // Create distributions 82 std::uniform_int_distribution<int> dist_batch(0, shape[3] - 1); 83 std::uniform_int_distribution<> dist_x1(0, scaled_width); 84 std::uniform_int_distribution<> dist_y1(0, scaled_height); 85 std::uniform_int_distribution<> dist_w(min_width, std::max(float(min_width), (pool_width - 2) * scaled_width)); 86 std::uniform_int_distribution<> dist_h(min_height, std::max(float(min_height), (pool_height - 2) * scaled_height)); 87 88 for(unsigned int pw = 0; pw < num_rois; ++pw) 89 { 90 const auto batch_idx = dist_batch(gen); 91 const auto x1 = dist_x1(gen); 92 const auto y1 = dist_y1(gen); 93 const auto x2 = x1 + dist_w(gen); 94 const auto y2 = y1 + dist_h(gen); 95 96 rois_ptr[values_per_roi * pw] = batch_idx; 97 rois_ptr[values_per_roi * pw + 1] = static_cast<uint16_t>(x1); 98 rois_ptr[values_per_roi * pw + 2] = static_cast<uint16_t>(y1); 99 rois_ptr[values_per_roi * pw + 3] = static_cast<uint16_t>(x2); 100 rois_ptr[values_per_roi * pw + 4] = static_cast<uint16_t>(y2); 101 } 102 } 103 compute_target(TensorShape input_shape,DataType data_type,DataLayout data_layout,const ROIPoolingLayerInfo & pool_info,const TensorShape rois_shape,const QuantizationInfo & qinfo,const QuantizationInfo & output_qinfo)104 TensorType compute_target(TensorShape input_shape, 105 DataType data_type, 106 DataLayout data_layout, 107 const ROIPoolingLayerInfo &pool_info, 108 const TensorShape rois_shape, 109 const QuantizationInfo &qinfo, 110 const QuantizationInfo &output_qinfo) 111 { 112 const QuantizationInfo rois_qinfo = is_data_type_quantized(data_type) ? QuantizationInfo(0.125f, 0) : QuantizationInfo(); 113 114 // Create tensors 115 TensorType src = create_tensor<TensorType>(input_shape, data_type, 1, qinfo, data_layout); 116 TensorType rois_tensor = create_tensor<TensorType>(rois_shape, _rois_data_type, 1, rois_qinfo); 117 118 // Initialise shape and declare output tensor dst 119 const TensorShape dst_shape; 120 TensorType dst = create_tensor<TensorType>(dst_shape, data_type, 1, output_qinfo, data_layout); 121 122 // Create and configure function 123 FunctionType roi_pool_layer; 124 roi_pool_layer.configure(&src, &rois_tensor, &dst, pool_info); 125 126 ARM_COMPUTE_ASSERT(src.info()->is_resizable()); 127 ARM_COMPUTE_ASSERT(rois_tensor.info()->is_resizable()); 128 ARM_COMPUTE_ASSERT(dst.info()->is_resizable()); 129 130 // Allocate tensors 131 src.allocator()->allocate(); 132 rois_tensor.allocator()->allocate(); 133 dst.allocator()->allocate(); 134 135 ARM_COMPUTE_ASSERT(!src.info()->is_resizable()); 136 ARM_COMPUTE_ASSERT(!rois_tensor.info()->is_resizable()); 137 ARM_COMPUTE_ASSERT(!dst.info()->is_resizable()); 138 139 // Fill tensors 140 fill(AccessorType(src)); 141 generate_rois(AccessorType(rois_tensor), input_shape, pool_info, rois_shape, data_layout); 142 143 // Compute function 144 roi_pool_layer.run(); 145 146 return dst; 147 } 148 compute_reference(const TensorShape & input_shape,DataType data_type,const ROIPoolingLayerInfo & pool_info,const TensorShape rois_shape,const QuantizationInfo & qinfo,const QuantizationInfo & output_qinfo)149 SimpleTensor<T> compute_reference(const TensorShape &input_shape, 150 DataType data_type, 151 const ROIPoolingLayerInfo &pool_info, 152 const TensorShape rois_shape, 153 const QuantizationInfo &qinfo, 154 const QuantizationInfo &output_qinfo) 155 { 156 // Create reference tensor 157 SimpleTensor<T> src{ input_shape, data_type, 1, qinfo }; 158 const QuantizationInfo rois_qinfo = is_data_type_quantized(data_type) ? QuantizationInfo(0.125f, 0) : QuantizationInfo(); 159 SimpleTensor<uint16_t> rois_tensor{ rois_shape, _rois_data_type, 1, rois_qinfo }; 160 161 // Fill reference tensor 162 fill(src); 163 generate_rois(rois_tensor, input_shape, pool_info, rois_shape); 164 165 return reference::roi_pool_layer(src, rois_tensor, pool_info, output_qinfo); 166 } 167 168 TensorType _target{}; 169 SimpleTensor<T> _reference{}; 170 const DataType _rois_data_type{ DataType::U16 }; 171 }; 172 173 template <typename TensorType, typename AccessorType, typename FunctionType, typename T> 174 class ROIPoolingLayerQuantizedFixture : public ROIPoolingLayerGenericFixture<TensorType, AccessorType, FunctionType, T> 175 { 176 public: 177 template <typename...> setup(TensorShape input_shape,const ROIPoolingLayerInfo pool_info,TensorShape rois_shape,DataType data_type,DataLayout data_layout,QuantizationInfo qinfo,QuantizationInfo output_qinfo)178 void setup(TensorShape input_shape, const ROIPoolingLayerInfo pool_info, TensorShape rois_shape, DataType data_type, 179 DataLayout data_layout, QuantizationInfo qinfo, QuantizationInfo output_qinfo) 180 { 181 ROIPoolingLayerGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(input_shape, pool_info, rois_shape, 182 data_type, data_layout, qinfo, output_qinfo); 183 } 184 }; 185 186 template <typename TensorType, typename AccessorType, typename FunctionType, typename T> 187 class ROIPoolingLayerFixture : public ROIPoolingLayerGenericFixture<TensorType, AccessorType, FunctionType, T> 188 { 189 public: 190 template <typename...> setup(TensorShape input_shape,const ROIPoolingLayerInfo pool_info,TensorShape rois_shape,DataType data_type,DataLayout data_layout)191 void setup(TensorShape input_shape, const ROIPoolingLayerInfo pool_info, TensorShape rois_shape, DataType data_type, DataLayout data_layout) 192 { 193 ROIPoolingLayerGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(input_shape, pool_info, rois_shape, data_type, data_layout, 194 QuantizationInfo(), QuantizationInfo()); 195 } 196 }; 197 198 } // namespace validation 199 } // namespace test 200 } // namespace arm_compute 201 202 #endif /* ARM_COMPUTE_TEST_ROIPOOLINGLAYER_FIXTURE */