Gitiles
Code Review Sign In
review.mlplatform.org / ml / ComputeLibrary / 575c5f1ae6bf6703b694d51d0818926809c7d8be / . / examples / graph_yolov3.cpp
blob: 5c8d3426ec78c3d4d9dc264d4240fb034e0ab2f9 [file] [log] [blame]
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]1/*
SiCong Li4841c972021-02-03 12:17:35 +0000[diff] [blame]2 * Copyright (c) 2018-2021 Arm Limited.
Michalis Spyrou177a9a52018-09-06 15:10:22 +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/graph.h"
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]25
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]26#include "support/ToolchainSupport.h"
27#include "utils/CommonGraphOptions.h"
28#include "utils/GraphUtils.h"
29#include "utils/Utils.h"
30
31using namespace arm_compute::utils;
32using namespace arm_compute::graph::frontend;
33using namespace arm_compute::graph_utils;
34
Georgios Pinitas108ab0b2018-09-14 18:35:11 +0100[diff] [blame]35/** Example demonstrating how to implement YOLOv3 network using the Compute Library's graph API */
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]36class GraphYOLOv3Example : public Example
37{
38public:
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]39 GraphYOLOv3Example() : cmd_parser(), common_opts(cmd_parser), common_params(), graph(0, "YOLOv3")
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]40 {
41 }
Michalis Spyroue22aa132018-09-13 10:35:33 +0100[diff] [blame]42
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]43 bool do_setup(int argc, char **argv) override
44 {
45 // Parse arguments
46 cmd_parser.parse(argc, argv);
Georgios Pinitascd60a5f2019-08-21 17:06:54 +0100[diff] [blame]47 cmd_parser.validate();
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]48
49 // Consume common parameters
50 common_params = consume_common_graph_parameters(common_opts);
51
52 // Return when help menu is requested
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]53 if (common_params.help)
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]54 {
55 cmd_parser.print_help(argv[0]);
56 return false;
57 }
58
59 // Checks
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]60 ARM_COMPUTE_EXIT_ON_MSG(arm_compute::is_data_type_quantized_asymmetric(common_params.data_type),
61 "QASYMM8 not supported for this graph");
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]62
63 // Print parameter values
64 std::cout << common_params << std::endl;
65
66 // Get trainable parameters data path
67 std::string data_path = common_params.data_path;
68
69 // Create a preprocessor object
Georgios Pinitas40f51a62020-11-21 03:04:18 +0000[diff] [blame]70 std::unique_ptr<IPreprocessor> preprocessor = std::make_unique<TFPreproccessor>(0.f);
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]71
72 // Create input descriptor
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]73 const TensorShape tensor_shape =
74 permute_shape(TensorShape(608U, 608U, 3U, 1U), DataLayout::NCHW, common_params.data_layout);
75 TensorDescriptor input_descriptor =
76 TensorDescriptor(tensor_shape, common_params.data_type).set_layout(common_params.data_layout);
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]77
78 // Set weights trained layout
79 const DataLayout weights_layout = DataLayout::NCHW;
80
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]81 graph << common_params.target << common_params.fast_math_hint
Michalis Spyroue22aa132018-09-13 10:35:33 +0100[diff] [blame]82 << InputLayer(input_descriptor, get_input_accessor(common_params, std::move(preprocessor), false));
83 std::pair<SubStream, SubStream> intermediate_layers = darknet53(data_path, weights_layout);
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]84 graph
85 << ConvolutionLayer(
86 1U, 1U, 512U,
87 get_weights_accessor(data_path, "/cnn_data/yolov3_model/conv2d_53_w.npy", weights_layout),
88 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 0, 0))
89 .set_name("conv2d_53")
90 << BatchNormalizationLayer(
91 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_53_mean.npy"),
92 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_53_var.npy"),
93 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_53_gamma.npy"),
94 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_53_beta.npy"), 0.000001f)
95 .set_name("conv2d_53/BatchNorm")
96 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LEAKY_RELU, 0.1f))
97 .set_name("conv2d_53/LeakyRelu")
98 << ConvolutionLayer(
99 3U, 3U, 1024U,
100 get_weights_accessor(data_path, "/cnn_data/yolov3_model/conv2d_54_w.npy", weights_layout),
101 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 1, 1))
102 .set_name("conv2d_54")
103 << BatchNormalizationLayer(
104 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_54_mean.npy"),
105 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_54_var.npy"),
106 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_54_gamma.npy"),
107 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_54_beta.npy"), 0.000001f)
108 .set_name("conv2d_54/BatchNorm")
109 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LEAKY_RELU, 0.1f))
110 .set_name("conv2d_54/LeakyRelu")
111 << ConvolutionLayer(
112 1U, 1U, 512U,
113 get_weights_accessor(data_path, "/cnn_data/yolov3_model/conv2d_55_w.npy", weights_layout),
114 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 0, 0))
115 .set_name("conv2d_55")
116 << BatchNormalizationLayer(
117 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_55_mean.npy"),
118 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_55_var.npy"),
119 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_55_gamma.npy"),
120 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_55_beta.npy"), 0.000001f)
121 .set_name("conv2d_55/BatchNorm")
122 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LEAKY_RELU, 0.1f))
123 .set_name("conv2d_55/LeakyRelu")
124 << ConvolutionLayer(
125 3U, 3U, 1024U,
126 get_weights_accessor(data_path, "/cnn_data/yolov3_model/conv2d_56_w.npy", weights_layout),
127 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 1, 1))
128 .set_name("conv2d_56")
129 << BatchNormalizationLayer(
130 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_56_mean.npy"),
131 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_56_var.npy"),
132 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_56_gamma.npy"),
133 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_56_beta.npy"), 0.000001f)
134 .set_name("conv2d_56/BatchNorm")
135 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LEAKY_RELU, 0.1f))
136 .set_name("conv2d_56/LeakyRelu")
137 << ConvolutionLayer(
138 1U, 1U, 512U,
139 get_weights_accessor(data_path, "/cnn_data/yolov3_model/conv2d_57_w.npy", weights_layout),
140 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 0, 0))
141 .set_name("conv2d_57")
142 << BatchNormalizationLayer(
143 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_57_mean.npy"),
144 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_57_var.npy"),
145 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_57_gamma.npy"),
146 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_57_beta.npy"), 0.000001f)
147 .set_name("conv2d_57/BatchNorm")
148 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LEAKY_RELU, 0.1f))
149 .set_name("conv2d_57/LeakyRelu");
Michalis Spyroue22aa132018-09-13 10:35:33 +0100[diff] [blame]150 SubStream route_1(graph);
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]151 graph
152 << ConvolutionLayer(
153 3U, 3U, 1024U,
154 get_weights_accessor(data_path, "/cnn_data/yolov3_model/conv2d_58_w.npy", weights_layout),
155 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 1, 1))
156 .set_name("conv2d_58")
157 << BatchNormalizationLayer(
158 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_58_mean.npy"),
159 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_58_var.npy"),
160 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_58_gamma.npy"),
161 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_58_beta.npy"), 0.000001f)
162 .set_name("conv2d_58/BatchNorm")
163 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LEAKY_RELU, 0.1f))
164 .set_name("conv2d_58/LeakyRelu")
165 << ConvolutionLayer(
166 1U, 1U, 255U,
167 get_weights_accessor(data_path, "/cnn_data/yolov3_model/conv2d_59_w.npy", weights_layout),
168 get_weights_accessor(data_path, "/cnn_data/yolov3_model/conv2d_59_b.npy", weights_layout),
169 PadStrideInfo(1, 1, 0, 0))
170 .set_name("conv2d_59")
171 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LINEAR, 1.f))
172 .set_name("conv2d_59/Linear")
173 << YOLOLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LOGISTIC, 0.1f)).set_name("Yolo1")
174 << OutputLayer(get_output_accessor(common_params, 5));
Michalis Spyroue22aa132018-09-13 10:35:33 +0100[diff] [blame]175 route_1 << ConvolutionLayer(
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]176 1U, 1U, 256U,
177 get_weights_accessor(data_path, "/cnn_data/yolov3_model/conv2d_60_w.npy", weights_layout),
178 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 0, 0))
179 .set_name("conv2d_60")
Michalis Spyroue22aa132018-09-13 10:35:33 +0100[diff] [blame]180 << BatchNormalizationLayer(
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]181 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_59_mean.npy"),
182 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_59_var.npy"),
183 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_59_gamma.npy"),
184 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_59_beta.npy"),
185 0.000001f)
186 .set_name("conv2d_59/BatchNorm")
187 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LEAKY_RELU, 0.1f))
188 .set_name("conv2d_60/LeakyRelu")
Georgios Pinitasc53266e2020-12-09 03:11:53 +0000[diff] [blame]189 << ResizeLayer(InterpolationPolicy::NEAREST_NEIGHBOR, 2, 2).set_name("Upsample_60");
Michalis Spyroue22aa132018-09-13 10:35:33 +0100[diff] [blame]190 SubStream concat_1(route_1);
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]191 concat_1
192 << ConcatLayer(std::move(route_1), std::move(intermediate_layers.second)).set_name("Route1")
193 << ConvolutionLayer(
194 1U, 1U, 256U,
195 get_weights_accessor(data_path, "/cnn_data/yolov3_model/conv2d_61_w.npy", weights_layout),
196 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 0, 0))
197 .set_name("conv2d_61")
198 << BatchNormalizationLayer(
199 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_60_mean.npy"),
200 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_60_var.npy"),
201 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_60_gamma.npy"),
202 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_60_beta.npy"), 0.000001f)
203 .set_name("conv2d_60/BatchNorm")
204 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LEAKY_RELU, 0.1f))
205 .set_name("conv2d_61/LeakyRelu")
206 << ConvolutionLayer(
207 3U, 3U, 512U,
208 get_weights_accessor(data_path, "/cnn_data/yolov3_model/conv2d_62_w.npy", weights_layout),
209 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 1, 1))
210 .set_name("conv2d_62")
211 << BatchNormalizationLayer(
212 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_61_mean.npy"),
213 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_61_var.npy"),
214 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_61_gamma.npy"),
215 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_61_beta.npy"), 0.000001f)
216 .set_name("conv2d_61/BatchNorm")
217 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LEAKY_RELU, 0.1f))
218 .set_name("conv2d_62/LeakyRelu")
219 << ConvolutionLayer(
220 1U, 1U, 256U,
221 get_weights_accessor(data_path, "/cnn_data/yolov3_model/conv2d_63_w.npy", weights_layout),
222 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 0, 0))
223 .set_name("conv2d_63")
224 << BatchNormalizationLayer(
225 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_62_mean.npy"),
226 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_62_var.npy"),
227 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_62_gamma.npy"),
228 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_62_beta.npy"), 0.000001f)
229 .set_name("conv2d_62/BatchNorm")
230 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LEAKY_RELU, 0.1f))
231 .set_name("conv2d_63/LeakyRelu")
232 << ConvolutionLayer(
233 3U, 3U, 512U,
234 get_weights_accessor(data_path, "/cnn_data/yolov3_model/conv2d_64_w.npy", weights_layout),
235 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 1, 1))
236 .set_name("conv2d_64")
237 << BatchNormalizationLayer(
238 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_63_mean.npy"),
239 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_63_var.npy"),
240 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_63_gamma.npy"),
241 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_63_beta.npy"), 0.000001f)
242 .set_name("conv2d_63/BatchNorm")
243 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LEAKY_RELU, 0.1f))
244 .set_name("conv2d_64/LeakyRelu")
245 << ConvolutionLayer(
246 1U, 1U, 256U,
247 get_weights_accessor(data_path, "/cnn_data/yolov3_model/conv2d_65_w.npy", weights_layout),
248 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 0, 0))
249 .set_name("conv2d_65")
250 << BatchNormalizationLayer(
251 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_64_mean.npy"),
252 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_64_var.npy"),
253 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_64_gamma.npy"),
254 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_64_beta.npy"), 0.000001f)
255 .set_name("conv2d_65/BatchNorm")
256 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LEAKY_RELU, 0.1f))
257 .set_name("conv2d_65/LeakyRelu");
Michalis Spyroue22aa132018-09-13 10:35:33 +0100[diff] [blame]258 SubStream route_2(concat_1);
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]259 concat_1
260 << ConvolutionLayer(
261 3U, 3U, 512U,
262 get_weights_accessor(data_path, "/cnn_data/yolov3_model/conv2d_66_w.npy", weights_layout),
263 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 1, 1))
264 .set_name("conv2d_66")
265 << BatchNormalizationLayer(
266 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_65_mean.npy"),
267 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_65_var.npy"),
268 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_65_gamma.npy"),
269 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_65_beta.npy"), 0.000001f)
270 .set_name("conv2d_65/BatchNorm")
271 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LEAKY_RELU, 0.1f))
272 .set_name("conv2d_66/LeakyRelu")
273 << ConvolutionLayer(
274 1U, 1U, 255U,
275 get_weights_accessor(data_path, "/cnn_data/yolov3_model/conv2d_67_w.npy", weights_layout),
276 get_weights_accessor(data_path, "/cnn_data/yolov3_model/conv2d_67_b.npy", weights_layout),
277 PadStrideInfo(1, 1, 0, 0))
278 .set_name("conv2d_67")
279 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LINEAR, 1.f))
280 .set_name("conv2d_67/Linear")
281 << YOLOLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LOGISTIC, 0.1f)).set_name("Yolo2")
282 << OutputLayer(get_output_accessor(common_params, 5));
Michalis Spyroue22aa132018-09-13 10:35:33 +0100[diff] [blame]283 route_2 << ConvolutionLayer(
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]284 1U, 1U, 128U,
285 get_weights_accessor(data_path, "/cnn_data/yolov3_model/conv2d_68_w.npy", weights_layout),
286 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 0, 0))
287 .set_name("conv2d_68")
Michalis Spyroue22aa132018-09-13 10:35:33 +0100[diff] [blame]288 << BatchNormalizationLayer(
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]289 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_66_mean.npy"),
290 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_66_var.npy"),
291 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_66_gamma.npy"),
292 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_66_beta.npy"),
293 0.000001f)
294 .set_name("conv2d_66/BatchNorm")
295 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LEAKY_RELU, 0.1f))
296 .set_name("conv2d_68/LeakyRelu")
Georgios Pinitasc53266e2020-12-09 03:11:53 +0000[diff] [blame]297 << ResizeLayer(InterpolationPolicy::NEAREST_NEIGHBOR, 2, 2).set_name("Upsample_68");
Michalis Spyroue22aa132018-09-13 10:35:33 +0100[diff] [blame]298 SubStream concat_2(route_2);
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]299 concat_2
300 << ConcatLayer(std::move(route_2), std::move(intermediate_layers.first)).set_name("Route2")
301 << ConvolutionLayer(
302 1U, 1U, 128U,
303 get_weights_accessor(data_path, "/cnn_data/yolov3_model/conv2d_69_w.npy", weights_layout),
304 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 0, 0))
305 .set_name("conv2d_69")
306 << BatchNormalizationLayer(
307 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_67_mean.npy"),
308 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_67_var.npy"),
309 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_67_gamma.npy"),
310 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_67_beta.npy"), 0.000001f)
311 .set_name("conv2d_67/BatchNorm")
312 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LEAKY_RELU, 0.1f))
313 .set_name("conv2d_69/LeakyRelu")
314 << ConvolutionLayer(
315 3U, 3U, 256U,
316 get_weights_accessor(data_path, "/cnn_data/yolov3_model/conv2d_70_w.npy", weights_layout),
317 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 1, 1))
318 .set_name("conv2d_70")
319 << BatchNormalizationLayer(
320 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_68_mean.npy"),
321 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_68_var.npy"),
322 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_68_gamma.npy"),
323 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_68_beta.npy"), 0.000001f)
324 .set_name("conv2d_68/BatchNorm")
325 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LEAKY_RELU, 0.1f))
326 .set_name("conv2d_70/LeakyRelu")
327 << ConvolutionLayer(
328 1U, 1U, 128U,
329 get_weights_accessor(data_path, "/cnn_data/yolov3_model/conv2d_71_w.npy", weights_layout),
330 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 0, 0))
331 .set_name("conv2d_71")
332 << BatchNormalizationLayer(
333 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_69_mean.npy"),
334 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_69_var.npy"),
335 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_69_gamma.npy"),
336 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_69_beta.npy"), 0.000001f)
337 .set_name("conv2d_69/BatchNorm")
338 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LEAKY_RELU, 0.1f))
339 .set_name("conv2d_71/LeakyRelu")
340 << ConvolutionLayer(
341 3U, 3U, 256U,
342 get_weights_accessor(data_path, "/cnn_data/yolov3_model/conv2d_72_w.npy", weights_layout),
343 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 1, 1))
344 .set_name("conv2d_72")
345 << BatchNormalizationLayer(
346 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_70_mean.npy"),
347 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_70_var.npy"),
348 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_70_gamma.npy"),
349 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_70_beta.npy"), 0.000001f)
350 .set_name("conv2d_70/BatchNorm")
351 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LEAKY_RELU, 0.1f))
352 .set_name("conv2d_72/LeakyRelu")
353 << ConvolutionLayer(
354 1U, 1U, 128U,
355 get_weights_accessor(data_path, "/cnn_data/yolov3_model/conv2d_73_w.npy", weights_layout),
356 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 0, 0))
357 .set_name("conv2d_73")
358 << BatchNormalizationLayer(
359 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_71_mean.npy"),
360 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_71_var.npy"),
361 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_71_gamma.npy"),
362 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_71_beta.npy"), 0.000001f)
363 .set_name("conv2d_71/BatchNorm")
364 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LEAKY_RELU, 0.1f))
365 .set_name("conv2d_73/LeakyRelu")
366 << ConvolutionLayer(
367 3U, 3U, 256U,
368 get_weights_accessor(data_path, "/cnn_data/yolov3_model/conv2d_74_w.npy", weights_layout),
369 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 1, 1))
370 .set_name("conv2d_74")
371 << BatchNormalizationLayer(
372 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_72_mean.npy"),
373 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_72_var.npy"),
374 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_72_gamma.npy"),
375 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_72_beta.npy"), 0.000001f)
376 .set_name("conv2d_72/BatchNorm")
377 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LEAKY_RELU, 0.1f))
378 .set_name("conv2d_74/LeakyRelu")
379 << ConvolutionLayer(
380 1U, 1U, 255U,
381 get_weights_accessor(data_path, "/cnn_data/yolov3_model/conv2d_75_w.npy", weights_layout),
382 get_weights_accessor(data_path, "/cnn_data/yolov3_model/conv2d_75_b.npy", weights_layout),
383 PadStrideInfo(1, 1, 0, 0))
384 .set_name("conv2d_75")
385 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LINEAR, 1.f))
386 .set_name("conv2d_75/Linear")
387 << YOLOLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LOGISTIC, 0.1f)).set_name("Yolo3")
388 << OutputLayer(get_output_accessor(common_params, 5));
Michalis Spyroue22aa132018-09-13 10:35:33 +0100[diff] [blame]389
390 // Finalize graph
391 GraphConfig config;
392 config.num_threads = common_params.threads;
393 config.use_tuner = common_params.enable_tuner;
Vidhya Sudhan Loganathan050471e2019-04-25 09:27:24 +0100[diff] [blame]394 config.tuner_mode = common_params.tuner_mode;
Michalis Spyroue22aa132018-09-13 10:35:33 +0100[diff] [blame]395 config.tuner_file = common_params.tuner_file;
SiCong Li4841c972021-02-03 12:17:35 +0000[diff] [blame]396 config.mlgo_file = common_params.mlgo_file;
Michalis Spyroue22aa132018-09-13 10:35:33 +0100[diff] [blame]397
398 graph.finalize(common_params.target, config);
399
400 return true;
401 }
402 void do_run() override
403 {
404 // Run graph
405 graph.run();
406 }
407
408private:
409 CommandLineParser cmd_parser;
410 CommonGraphOptions common_opts;
411 CommonGraphParams common_params;
412 Stream graph;
413
414 std::pair<SubStream, SubStream> darknet53(const std::string &data_path, DataLayout weights_layout)
415 {
416 graph << ConvolutionLayer(
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]417 3U, 3U, 32U,
418 get_weights_accessor(data_path, "/cnn_data/yolov3_model/conv2d_1_w.npy", weights_layout),
419 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 1, 1))
420 .set_name("conv2d_1/Conv2D")
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]421 << BatchNormalizationLayer(
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]422 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_1_mean.npy"),
423 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_1_var.npy"),
424 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_1_gamma.npy"),
425 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_1_beta.npy"),
426 0.000001f)
427 .set_name("conv2d_1/BatchNorm")
428 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LEAKY_RELU, 0.1f))
429 .set_name("conv2d_1/LeakyRelu")
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]430 << ConvolutionLayer(
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]431 3U, 3U, 64U,
432 get_weights_accessor(data_path, "/cnn_data/yolov3_model/conv2d_2_w.npy", weights_layout),
433 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(2, 2, 1, 1))
434 .set_name("conv2d_2/Conv2D")
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]435 << BatchNormalizationLayer(
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]436 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_2_mean.npy"),
437 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_2_var.npy"),
438 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_2_gamma.npy"),
439 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_2_beta.npy"),
440 0.000001f)
441 .set_name("conv2d_2/BatchNorm")
442 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LEAKY_RELU, 0.1f))
443 .set_name("conv2d_2/LeakyRelu");
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]444 darknet53_block(data_path, "3", weights_layout, 32U);
445 graph << ConvolutionLayer(
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]446 3U, 3U, 128U,
447 get_weights_accessor(data_path, "/cnn_data/yolov3_model/conv2d_5_w.npy", weights_layout),
448 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(2, 2, 1, 1))
449 .set_name("conv2d_5/Conv2D")
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]450 << BatchNormalizationLayer(
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]451 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_5_mean.npy"),
452 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_5_var.npy"),
453 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_5_gamma.npy"),
454 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_5_beta.npy"),
455 0.000001f)
456 .set_name("conv2d_5/BatchNorm")
457 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LEAKY_RELU, 0.1f))
458 .set_name("conv2d_5/LeakyRelu");
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]459 darknet53_block(data_path, "6", weights_layout, 64U);
460 darknet53_block(data_path, "8", weights_layout, 64U);
461 graph << ConvolutionLayer(
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]462 3U, 3U, 256U,
463 get_weights_accessor(data_path, "/cnn_data/yolov3_model/conv2d_10_w.npy", weights_layout),
464 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(2, 2, 1, 1))
465 .set_name("conv2d_10/Conv2D")
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]466 << BatchNormalizationLayer(
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]467 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_10_mean.npy"),
468 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_10_var.npy"),
469 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_10_gamma.npy"),
470 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_10_beta.npy"),
471 0.000001f)
472 .set_name("conv2d_10/BatchNorm")
473 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LEAKY_RELU, 0.1f))
474 .set_name("conv2d_10/LeakyRelu");
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]475 darknet53_block(data_path, "11", weights_layout, 128U);
476 darknet53_block(data_path, "13", weights_layout, 128U);
477 darknet53_block(data_path, "15", weights_layout, 128U);
478 darknet53_block(data_path, "17", weights_layout, 128U);
479 darknet53_block(data_path, "19", weights_layout, 128U);
480 darknet53_block(data_path, "21", weights_layout, 128U);
481 darknet53_block(data_path, "23", weights_layout, 128U);
482 darknet53_block(data_path, "25", weights_layout, 128U);
Michalis Spyroue22aa132018-09-13 10:35:33 +0100[diff] [blame]483 SubStream layer_36(graph);
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]484 graph << ConvolutionLayer(
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]485 3U, 3U, 512U,
486 get_weights_accessor(data_path, "/cnn_data/yolov3_model/conv2d_27_w.npy", weights_layout),
487 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(2, 2, 1, 1))
488 .set_name("conv2d_27/Conv2D")
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]489 << BatchNormalizationLayer(
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]490 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_27_mean.npy"),
491 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_27_var.npy"),
492 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_27_gamma.npy"),
493 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_27_beta.npy"),
494 0.000001f)
495 .set_name("conv2d_27/BatchNorm")
496 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LEAKY_RELU, 0.1f))
497 .set_name("conv2d_27/LeakyRelu");
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]498 darknet53_block(data_path, "28", weights_layout, 256U);
499 darknet53_block(data_path, "30", weights_layout, 256U);
500 darknet53_block(data_path, "32", weights_layout, 256U);
501 darknet53_block(data_path, "34", weights_layout, 256U);
502 darknet53_block(data_path, "36", weights_layout, 256U);
503 darknet53_block(data_path, "38", weights_layout, 256U);
504 darknet53_block(data_path, "40", weights_layout, 256U);
505 darknet53_block(data_path, "42", weights_layout, 256U);
Michalis Spyroue22aa132018-09-13 10:35:33 +0100[diff] [blame]506 SubStream layer_61(graph);
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]507 graph << ConvolutionLayer(
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]508 3U, 3U, 1024U,
509 get_weights_accessor(data_path, "/cnn_data/yolov3_model/conv2d_44_w.npy", weights_layout),
510 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(2, 2, 1, 1))
511 .set_name("conv2d_44/Conv2D")
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]512 << BatchNormalizationLayer(
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]513 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_44_mean.npy"),
514 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_44_var.npy"),
515 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_44_gamma.npy"),
516 get_weights_accessor(data_path, "/cnn_data/yolov3_model/batch_normalization_44_beta.npy"),
517 0.000001f)
518 .set_name("conv2d_44/BatchNorm")
519 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LEAKY_RELU, 0.1f))
520 .set_name("conv2d_44/LeakyRelu");
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]521 darknet53_block(data_path, "45", weights_layout, 512U);
522 darknet53_block(data_path, "47", weights_layout, 512U);
523 darknet53_block(data_path, "49", weights_layout, 512U);
524 darknet53_block(data_path, "51", weights_layout, 512U);
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]525
Michalis Spyroue22aa132018-09-13 10:35:33 +0100[diff] [blame]526 return std::pair<SubStream, SubStream>(layer_36, layer_61);
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]527 }
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]528
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]529 void darknet53_block(const std::string &data_path,
530 std::string &&param_path,
531 DataLayout weights_layout,
532 unsigned int filter_size)
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]533 {
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]534 std::string total_path = "/cnn_data/yolov3_model/";
535 std::string param_path2 =
536 arm_compute::support::cpp11::to_string(arm_compute::support::cpp11::stoi(param_path) + 1);
537 SubStream i_a(graph);
538 SubStream i_b(graph);
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]539 i_a << ConvolutionLayer(
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]540 1U, 1U, filter_size,
541 get_weights_accessor(data_path, total_path + "conv2d_" + param_path + "_w.npy", weights_layout),
542 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 0, 0))
543 .set_name("conv2d_" + param_path + "/Conv2D")
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]544 << BatchNormalizationLayer(
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]545 get_weights_accessor(data_path, total_path + "batch_normalization_" + param_path + "_mean.npy"),
546 get_weights_accessor(data_path, total_path + "batch_normalization_" + param_path + "_var.npy"),
547 get_weights_accessor(data_path, total_path + "batch_normalization_" + param_path + "_gamma.npy"),
548 get_weights_accessor(data_path, total_path + "batch_normalization_" + param_path + "_beta.npy"),
549 0.000001f)
550 .set_name("conv2d_" + param_path + "/BatchNorm")
551 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LEAKY_RELU, 0.1f))
552 .set_name("conv2d_" + param_path + "/LeakyRelu")
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]553 << ConvolutionLayer(
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]554 3U, 3U, filter_size * 2,
555 get_weights_accessor(data_path, total_path + "conv2d_" + param_path2 + "_w.npy", weights_layout),
556 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 1, 1))
557 .set_name("conv2d_" + param_path2 + "/Conv2D")
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]558 << BatchNormalizationLayer(
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]559 get_weights_accessor(data_path, total_path + "batch_normalization_" + param_path2 + "_mean.npy"),
560 get_weights_accessor(data_path, total_path + "batch_normalization_" + param_path2 + "_var.npy"),
561 get_weights_accessor(data_path, total_path + "batch_normalization_" + param_path2 + "_gamma.npy"),
562 get_weights_accessor(data_path, total_path + "batch_normalization_" + param_path2 + "_beta.npy"),
563 0.000001f)
564 .set_name("conv2d_" + param_path2 + "/BatchNorm")
565 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LEAKY_RELU, 0.1f))
566 .set_name("conv2d_" + param_path2 + "/LeakyRelu");
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]567
Felix Thomasmathibalanafd38f02023-09-27 17:46:17 +0100[diff] [blame]568 graph << EltwiseLayer(std::move(i_a), std::move(i_b), EltwiseOperation::Add)
569 .set_name("")
570 .set_name("add_" + param_path + "_" + param_path2);
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]571 }
572};
573
574/** Main program for YOLOv3
575 *
Georgios Pinitasbdbbbe82018-11-07 16:06:47 +0000[diff] [blame]576 * Model is based on:
577 * https://arxiv.org/abs/1804.02767
578 * "YOLOv3: An Incremental Improvement"
579 * Joseph Redmon, Ali Farhadi
580 *
Michalis Spyrou177a9a52018-09-06 15:10:22 +0100[diff] [blame]581 * @note To list all the possible arguments execute the binary appended with the --help option
582 *
583 * @param[in] argc Number of arguments
584 * @param[in] argv Arguments
585 *
586 * @return Return code
587 */
588int main(int argc, char **argv)
589{
590 return arm_compute::utils::run_example<GraphYOLOv3Example>(argc, argv);
591}