1 /*
2 * Copyright (c) 2017-2020 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 #include "arm_compute/core/CPP/kernels/CPPDetectionWindowNonMaximaSuppressionKernel.h"
25
26 #include "arm_compute/core/Helpers.h"
27
28 #include <algorithm>
29 #include <cmath>
30
31 using namespace arm_compute;
32
33 namespace
34 {
compare_detection_window(const DetectionWindow & lhs,const DetectionWindow & rhs)35 bool compare_detection_window(const DetectionWindow &lhs, const DetectionWindow &rhs)
36 {
37 if(lhs.idx_class < rhs.idx_class)
38 {
39 return true;
40 }
41 if(rhs.idx_class < lhs.idx_class)
42 {
43 return false;
44 }
45
46 // idx_classes are equal so compare by score
47 if(lhs.score > rhs.score)
48 {
49 return true;
50 }
51 if(rhs.score > lhs.score)
52 {
53 return false;
54 }
55
56 return false;
57 }
58 } // namespace
59
CPPDetectionWindowNonMaximaSuppressionKernel()60 CPPDetectionWindowNonMaximaSuppressionKernel::CPPDetectionWindowNonMaximaSuppressionKernel()
61 : _input_output(nullptr), _min_distance(0.0f)
62 {
63 }
64
is_parallelisable() const65 bool CPPDetectionWindowNonMaximaSuppressionKernel::is_parallelisable() const
66 {
67 return false;
68 }
69
configure(IDetectionWindowArray * input_output,float min_distance)70 void CPPDetectionWindowNonMaximaSuppressionKernel::configure(IDetectionWindowArray *input_output, float min_distance)
71 {
72 ARM_COMPUTE_ERROR_ON(nullptr == input_output);
73
74 _input_output = input_output;
75 _min_distance = min_distance;
76
77 IKernel::configure(Window()); // Default 1 iteration window
78 }
79
run(const Window & window,const ThreadInfo & info)80 void CPPDetectionWindowNonMaximaSuppressionKernel::run(const Window &window, const ThreadInfo &info)
81 {
82 ARM_COMPUTE_UNUSED(info);
83 ARM_COMPUTE_UNUSED(window);
84 ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
85 ARM_COMPUTE_ERROR_ON_MISMATCHING_WINDOWS(IKernel::window(), window);
86 ARM_COMPUTE_ERROR_ON(_input_output->buffer() == nullptr);
87
88 const size_t num_candidates = _input_output->num_values();
89 size_t num_detections = 0;
90
91 // Sort list of candidates by idx_class and then score
92 std::sort(_input_output->buffer(), _input_output->buffer() + num_candidates, compare_detection_window);
93
94 const float min_distance_pow2 = _min_distance * _min_distance;
95
96 // Euclidean distance
97 for(size_t i = 0; i < num_candidates; ++i)
98 {
99 if(0.0f != _input_output->at(i).score)
100 {
101 DetectionWindow cur;
102 cur.x = _input_output->at(i).x;
103 cur.y = _input_output->at(i).y;
104 cur.width = _input_output->at(i).width;
105 cur.height = _input_output->at(i).height;
106 cur.idx_class = _input_output->at(i).idx_class;
107 cur.score = _input_output->at(i).score;
108
109 // Store window
110 _input_output->at(num_detections) = cur;
111
112 ++num_detections;
113
114 const float xc = cur.x + cur.width * 0.5f;
115 const float yc = cur.y + cur.height * 0.5f;
116
117 for(size_t k = i + 1; k < (num_candidates) && (cur.idx_class == _input_output->at(k).idx_class); ++k)
118 {
119 const float xn = _input_output->at(k).x + _input_output->at(k).width * 0.5f;
120 const float yn = _input_output->at(k).y + _input_output->at(k).height * 0.5f;
121
122 const float dx = std::fabs(xn - xc);
123 const float dy = std::fabs(yn - yc);
124
125 if(dx < _min_distance && dy < _min_distance)
126 {
127 const float d = dx * dx + dy * dy;
128
129 if(d < min_distance_pow2)
130 {
131 // Invalidate detection window
132 _input_output->at(k).score = 0.0f;
133 }
134 }
135 }
136 }
137 }
138
139 _input_output->resize(num_detections);
140 }
141