• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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 */