• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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/graph.h"
25 #include "support/ToolchainSupport.h"
26 #include "utils/CommonGraphOptions.h"
27 #include "utils/GraphUtils.h"
28 #include "utils/Utils.h"
29 
30 using namespace arm_compute;
31 using namespace arm_compute::utils;
32 using namespace arm_compute::graph::frontend;
33 using namespace arm_compute::graph_utils;
34 
35 /** Example demonstrating how to implement MobileNet's network using the Compute Library's graph API */
36 class GraphMobilenetExample : public Example
37 {
38 public:
GraphMobilenetExample()39     GraphMobilenetExample()
40         : cmd_parser(), common_opts(cmd_parser), common_params(), graph(0, "MobileNetV1")
41     {
42         // Add model id option
43         model_id_opt = cmd_parser.add_option<SimpleOption<int>>("model-id", 0);
44         model_id_opt->set_help("Mobilenet model id (0: 1.0_224, else: 0.75_160");
45     }
46     GraphMobilenetExample(const GraphMobilenetExample &) = delete;
47     GraphMobilenetExample &operator=(const GraphMobilenetExample &) = delete;
48     ~GraphMobilenetExample() override                               = default;
do_setup(int argc,char ** argv)49     bool do_setup(int argc, char **argv) override
50     {
51         // Parse arguments
52         cmd_parser.parse(argc, argv);
53         cmd_parser.validate();
54 
55         // Consume common parameters
56         common_params = consume_common_graph_parameters(common_opts);
57 
58         // Return when help menu is requested
59         if(common_params.help)
60         {
61             cmd_parser.print_help(argv[0]);
62             return false;
63         }
64 
65         // Print parameter values
66         std::cout << common_params << std::endl;
67 
68         // Get model parameters
69         int model_id = model_id_opt->value();
70 
71         // Create input descriptor
72         unsigned int spatial_size = (model_id == 0 || common_params.data_type == DataType::QASYMM8) ? 224 : 160;
73 
74         // Create input descriptor
75         const TensorShape tensor_shape     = permute_shape(TensorShape(spatial_size, spatial_size, 3U, 1U), DataLayout::NCHW, common_params.data_layout);
76         TensorDescriptor  input_descriptor = TensorDescriptor(tensor_shape, common_params.data_type).set_layout(common_params.data_layout);
77 
78         // Set graph hints
79         graph << common_params.target
80               << common_params.fast_math_hint;
81 
82         // Create core graph
83         if(arm_compute::is_data_type_float(common_params.data_type))
84         {
85             create_graph_float(input_descriptor, model_id);
86         }
87         else
88         {
89             create_graph_qasymm(input_descriptor);
90         }
91 
92         // Create common tail
93         graph << ReshapeLayer(TensorShape(1001U)).set_name("Reshape")
94               << SoftmaxLayer().set_name("Softmax")
95               << OutputLayer(get_output_accessor(common_params, 5));
96 
97         // Finalize graph
98         GraphConfig config;
99         config.num_threads = common_params.threads;
100         config.use_tuner   = common_params.enable_tuner;
101         config.tuner_mode  = common_params.tuner_mode;
102         config.tuner_file  = common_params.tuner_file;
103 
104         graph.finalize(common_params.target, config);
105 
106         return true;
107     }
do_run()108     void do_run() override
109     {
110         // Run graph
111         graph.run();
112     }
113 
114 private:
115     CommandLineParser  cmd_parser;
116     CommonGraphOptions common_opts;
117     SimpleOption<int> *model_id_opt{ nullptr };
118     CommonGraphParams  common_params;
119     Stream             graph;
120 
create_graph_float(TensorDescriptor & input_descriptor,int model_id)121     void create_graph_float(TensorDescriptor &input_descriptor, int model_id)
122     {
123         float       depth_scale = (model_id == 0) ? 1.f : 0.75;
124         std::string model_path  = (model_id == 0) ? "/cnn_data/mobilenet_v1_1_224_model/" : "/cnn_data/mobilenet_v1_075_160_model/";
125 
126         // Create a preprocessor object
127         std::unique_ptr<IPreprocessor> preprocessor = arm_compute::support::cpp14::make_unique<TFPreproccessor>();
128 
129         // Get trainable parameters data path
130         std::string data_path = common_params.data_path;
131 
132         // Add model path to data path
133         if(!data_path.empty())
134         {
135             data_path += model_path;
136         }
137 
138         graph << InputLayer(input_descriptor,
139                             get_input_accessor(common_params, std::move(preprocessor), false))
140               << ConvolutionLayer(
141                   3U, 3U, 32U * depth_scale,
142                   get_weights_accessor(data_path, "Conv2d_0_weights.npy", DataLayout::NCHW),
143                   std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
144                   PadStrideInfo(2, 2, 0, 1, 0, 1, DimensionRoundingType::FLOOR))
145               .set_name("Conv2d_0")
146               << BatchNormalizationLayer(
147                   get_weights_accessor(data_path, "Conv2d_0_BatchNorm_moving_mean.npy"),
148                   get_weights_accessor(data_path, "Conv2d_0_BatchNorm_moving_variance.npy"),
149                   get_weights_accessor(data_path, "Conv2d_0_BatchNorm_gamma.npy"),
150                   get_weights_accessor(data_path, "Conv2d_0_BatchNorm_beta.npy"),
151                   0.001f)
152               .set_name("Conv2d_0/BatchNorm")
153               << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::BOUNDED_RELU, 6.f)).set_name("Conv2d_0/Relu6");
154         graph << get_dwsc_node_float(data_path, "Conv2d_1", 64 * depth_scale, PadStrideInfo(1, 1, 1, 1), PadStrideInfo(1, 1, 0, 0));
155         graph << get_dwsc_node_float(data_path, "Conv2d_2", 128 * depth_scale, PadStrideInfo(2, 2, 0, 1, 0, 1, DimensionRoundingType::CEIL), PadStrideInfo(1, 1, 0, 0));
156         graph << get_dwsc_node_float(data_path, "Conv2d_3", 128 * depth_scale, PadStrideInfo(1, 1, 1, 1, 1, 1, DimensionRoundingType::CEIL), PadStrideInfo(1, 1, 0, 0));
157         graph << get_dwsc_node_float(data_path, "Conv2d_4", 256 * depth_scale, PadStrideInfo(2, 2, 0, 1, 0, 1, DimensionRoundingType::CEIL), PadStrideInfo(1, 1, 0, 0));
158         graph << get_dwsc_node_float(data_path, "Conv2d_5", 256 * depth_scale, PadStrideInfo(1, 1, 1, 1, 1, 1, DimensionRoundingType::CEIL), PadStrideInfo(1, 1, 0, 0));
159         graph << get_dwsc_node_float(data_path, "Conv2d_6", 512 * depth_scale, PadStrideInfo(2, 2, 0, 1, 0, 1, DimensionRoundingType::CEIL), PadStrideInfo(1, 1, 0, 0));
160         graph << get_dwsc_node_float(data_path, "Conv2d_7", 512 * depth_scale, PadStrideInfo(1, 1, 1, 1, 1, 1, DimensionRoundingType::CEIL), PadStrideInfo(1, 1, 0, 0));
161         graph << get_dwsc_node_float(data_path, "Conv2d_8", 512 * depth_scale, PadStrideInfo(1, 1, 1, 1, 1, 1, DimensionRoundingType::CEIL), PadStrideInfo(1, 1, 0, 0));
162         graph << get_dwsc_node_float(data_path, "Conv2d_9", 512 * depth_scale, PadStrideInfo(1, 1, 1, 1, 1, 1, DimensionRoundingType::CEIL), PadStrideInfo(1, 1, 0, 0));
163         graph << get_dwsc_node_float(data_path, "Conv2d_10", 512 * depth_scale, PadStrideInfo(1, 1, 1, 1, 1, 1, DimensionRoundingType::CEIL), PadStrideInfo(1, 1, 0, 0));
164         graph << get_dwsc_node_float(data_path, "Conv2d_11", 512 * depth_scale, PadStrideInfo(1, 1, 1, 1, 1, 1, DimensionRoundingType::CEIL), PadStrideInfo(1, 1, 0, 0));
165         graph << get_dwsc_node_float(data_path, "Conv2d_12", 1024 * depth_scale, PadStrideInfo(2, 2, 0, 1, 0, 1, DimensionRoundingType::CEIL), PadStrideInfo(1, 1, 0, 0));
166         graph << get_dwsc_node_float(data_path, "Conv2d_13", 1024 * depth_scale, PadStrideInfo(1, 1, 1, 1, 1, 1, DimensionRoundingType::CEIL), PadStrideInfo(1, 1, 0, 0));
167         graph << PoolingLayer(PoolingLayerInfo(PoolingType::AVG, common_params.data_layout)).set_name("Logits/AvgPool_1a")
168               << ConvolutionLayer(
169                   1U, 1U, 1001U,
170                   get_weights_accessor(data_path, "Logits_Conv2d_1c_1x1_weights.npy", DataLayout::NCHW),
171                   get_weights_accessor(data_path, "Logits_Conv2d_1c_1x1_biases.npy"),
172                   PadStrideInfo(1, 1, 0, 0))
173               .set_name("Logits/Conv2d_1c_1x1");
174     }
175 
create_graph_qasymm(TensorDescriptor & input_descriptor)176     void create_graph_qasymm(TensorDescriptor &input_descriptor)
177     {
178         // Get trainable parameters data path
179         std::string data_path = common_params.data_path;
180 
181         // Add model path to data path
182         if(!data_path.empty())
183         {
184             data_path += "/cnn_data/mobilenet_qasymm8_model/";
185         }
186 
187         // Quantization info taken from the AndroidNN QASYMM8 MobileNet example
188         const QuantizationInfo in_quant_info = QuantizationInfo(0.0078125f, 128);
189 
190         const std::vector<QuantizationInfo> conv_weights_quant_info =
191         {
192             QuantizationInfo(0.02182667888700962f, 151), // conv0
193             QuantizationInfo(0.004986600950360298f, 74)  // conv14
194         };
195         const std::vector<QuantizationInfo> conv_out_quant_info =
196         {
197             QuantizationInfo(0.023528477177023888f, 0), // conv0
198             QuantizationInfo(0.16609922051429749f, 66)  // conv14
199         };
200 
201         const std::vector<QuantizationInfo> depth_weights_quant_info =
202         {
203             QuantizationInfo(0.29219913482666016f, 110),  // dwsc1
204             QuantizationInfo(0.40277284383773804f, 130),  // dwsc2
205             QuantizationInfo(0.06053730100393295f, 160),  // dwsc3
206             QuantizationInfo(0.01675807684659958f, 123),  // dwsc4
207             QuantizationInfo(0.04105526953935623f, 129),  // dwsc5
208             QuantizationInfo(0.013460792601108551f, 122), // dwsc6
209             QuantizationInfo(0.036934755742549896f, 132), // dwsc7
210             QuantizationInfo(0.042609862983226776f, 94),  // dwsc8
211             QuantizationInfo(0.028358859941363335f, 127), // dwsc9
212             QuantizationInfo(0.024329448118805885f, 134), // dwsc10
213             QuantizationInfo(0.019366811960935593f, 106), // dwsc11
214             QuantizationInfo(0.007835594937205315f, 126), // dwsc12
215             QuantizationInfo(0.12616927921772003f, 211)   // dwsc13
216         };
217 
218         const std::vector<QuantizationInfo> point_weights_quant_info =
219         {
220             QuantizationInfo(0.030420949682593346f, 121), // dwsc1
221             QuantizationInfo(0.015148180536925793f, 104), // dwsc2
222             QuantizationInfo(0.013755458407104015f, 94),  // dwsc3
223             QuantizationInfo(0.007601846940815449f, 151), // dwsc4
224             QuantizationInfo(0.006431614048779011f, 122), // dwsc5
225             QuantizationInfo(0.00917122047394514f, 109),  // dwsc6
226             QuantizationInfo(0.005300046876072884f, 140), // dwsc7
227             QuantizationInfo(0.0049632852897048f, 127),   // dwsc8
228             QuantizationInfo(0.007770895957946777f, 89),  // dwsc9
229             QuantizationInfo(0.009658650495111942f, 99),  // dwsc10
230             QuantizationInfo(0.005446993745863438f, 153), // dwsc11
231             QuantizationInfo(0.00817922968417406f, 130),  // dwsc12
232             QuantizationInfo(0.018048152327537537f, 95)   // dwsc13
233         };
234 
235         graph << InputLayer(input_descriptor.set_quantization_info(in_quant_info),
236                             get_input_accessor(common_params, nullptr, false))
237               << ConvolutionLayer(
238                   3U, 3U, 32U,
239                   get_weights_accessor(data_path, "Conv2d_0_weights.npy"),
240                   get_weights_accessor(data_path, "Conv2d_0_bias.npy"),
241                   PadStrideInfo(2U, 2U, 0U, 1U, 0U, 1U, DimensionRoundingType::FLOOR),
242                   1, conv_weights_quant_info.at(0), conv_out_quant_info.at(0))
243               .set_name("Conv2d_0")
244               << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LU_BOUNDED_RELU, 6.f)).set_name("Conv2d_0/Relu6");
245         graph << get_dwsc_node_qasymm(data_path, "Conv2d_1", 64U, PadStrideInfo(1U, 1U, 1U, 1U), PadStrideInfo(1U, 1U, 0U, 0U), depth_weights_quant_info.at(0), point_weights_quant_info.at(0));
246         graph << get_dwsc_node_qasymm(data_path, "Conv2d_2", 128U, PadStrideInfo(2U, 2U, 0U, 1U, 0U, 1U, DimensionRoundingType::FLOOR), PadStrideInfo(1U, 1U, 0U, 0U), depth_weights_quant_info.at(1),
247                                       point_weights_quant_info.at(1));
248         graph << get_dwsc_node_qasymm(data_path, "Conv2d_3", 128U, PadStrideInfo(1U, 1U, 1U, 1U, 1U, 1U, DimensionRoundingType::FLOOR), PadStrideInfo(1U, 1U, 0U, 0U), depth_weights_quant_info.at(2),
249                                       point_weights_quant_info.at(2));
250         graph << get_dwsc_node_qasymm(data_path, "Conv2d_4", 256U, PadStrideInfo(2U, 2U, 0U, 1U, 0U, 1U, DimensionRoundingType::FLOOR), PadStrideInfo(1U, 1U, 0U, 0U), depth_weights_quant_info.at(3),
251                                       point_weights_quant_info.at(3));
252         graph << get_dwsc_node_qasymm(data_path, "Conv2d_5", 256U, PadStrideInfo(1U, 1U, 1U, 1U, 1U, 1U, DimensionRoundingType::FLOOR), PadStrideInfo(1U, 1U, 0U, 0U), depth_weights_quant_info.at(4),
253                                       point_weights_quant_info.at(4));
254         graph << get_dwsc_node_qasymm(data_path, "Conv2d_6", 512U, PadStrideInfo(2U, 2U, 0U, 1U, 0U, 1U, DimensionRoundingType::FLOOR), PadStrideInfo(1U, 1U, 0U, 0U), depth_weights_quant_info.at(5),
255                                       point_weights_quant_info.at(5));
256         graph << get_dwsc_node_qasymm(data_path, "Conv2d_7", 512U, PadStrideInfo(1U, 1U, 1U, 1U, 1U, 1U, DimensionRoundingType::FLOOR), PadStrideInfo(1U, 1U, 0U, 0U), depth_weights_quant_info.at(6),
257                                       point_weights_quant_info.at(6));
258         graph << get_dwsc_node_qasymm(data_path, "Conv2d_8", 512U, PadStrideInfo(1U, 1U, 1U, 1U, 1U, 1U, DimensionRoundingType::FLOOR), PadStrideInfo(1U, 1U, 0U, 0U), depth_weights_quant_info.at(7),
259                                       point_weights_quant_info.at(7));
260         graph << get_dwsc_node_qasymm(data_path, "Conv2d_9", 512U, PadStrideInfo(1U, 1U, 1U, 1U, 1U, 1U, DimensionRoundingType::FLOOR), PadStrideInfo(1U, 1U, 0U, 0U), depth_weights_quant_info.at(8),
261                                       point_weights_quant_info.at(8));
262         graph << get_dwsc_node_qasymm(data_path, "Conv2d_10", 512U, PadStrideInfo(1U, 1U, 1U, 1U, 1U, 1U, DimensionRoundingType::FLOOR), PadStrideInfo(1U, 1U, 0U, 0U), depth_weights_quant_info.at(9),
263                                       point_weights_quant_info.at(9));
264         graph << get_dwsc_node_qasymm(data_path, "Conv2d_11", 512U, PadStrideInfo(1U, 1U, 1U, 1U, 1U, 1U, DimensionRoundingType::FLOOR), PadStrideInfo(1U, 1U, 0U, 0U), depth_weights_quant_info.at(10),
265                                       point_weights_quant_info.at(10));
266         graph << get_dwsc_node_qasymm(data_path, "Conv2d_12", 1024U, PadStrideInfo(2U, 2U, 0U, 1U, 0U, 1U, DimensionRoundingType::FLOOR), PadStrideInfo(1U, 1U, 0U, 0U), depth_weights_quant_info.at(11),
267                                       point_weights_quant_info.at(11));
268         graph << get_dwsc_node_qasymm(data_path, "Conv2d_13", 1024U, PadStrideInfo(1U, 1U, 1U, 1U, 1U, 1U, DimensionRoundingType::FLOOR), PadStrideInfo(1U, 1U, 0U, 0U), depth_weights_quant_info.at(12),
269                                       point_weights_quant_info.at(12))
270               << PoolingLayer(PoolingLayerInfo(PoolingType::AVG, common_params.data_layout)).set_name("Logits/AvgPool_1a")
271               << ConvolutionLayer(
272                   1U, 1U, 1001U,
273                   get_weights_accessor(data_path, "Logits_Conv2d_1c_1x1_weights.npy"),
274                   get_weights_accessor(data_path, "Logits_Conv2d_1c_1x1_bias.npy"),
275                   PadStrideInfo(1U, 1U, 0U, 0U), 1, conv_weights_quant_info.at(1), conv_out_quant_info.at(1))
276               .set_name("Logits/Conv2d_1c_1x1");
277     }
278 
get_dwsc_node_float(const std::string & data_path,std::string && param_path,unsigned int conv_filt,PadStrideInfo dwc_pad_stride_info,PadStrideInfo conv_pad_stride_info)279     ConcatLayer get_dwsc_node_float(const std::string &data_path, std::string &&param_path,
280                                     unsigned int  conv_filt,
281                                     PadStrideInfo dwc_pad_stride_info, PadStrideInfo conv_pad_stride_info)
282     {
283         std::string total_path = param_path + "_";
284         SubStream   sg(graph);
285         sg << DepthwiseConvolutionLayer(
286                3U, 3U,
287                get_weights_accessor(data_path, total_path + "depthwise_depthwise_weights.npy", DataLayout::NCHW),
288                std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
289                dwc_pad_stride_info)
290            .set_name(total_path + "depthwise/depthwise")
291            << BatchNormalizationLayer(
292                get_weights_accessor(data_path, total_path + "depthwise_BatchNorm_moving_mean.npy"),
293                get_weights_accessor(data_path, total_path + "depthwise_BatchNorm_moving_variance.npy"),
294                get_weights_accessor(data_path, total_path + "depthwise_BatchNorm_gamma.npy"),
295                get_weights_accessor(data_path, total_path + "depthwise_BatchNorm_beta.npy"),
296                0.001f)
297            .set_name(total_path + "depthwise/BatchNorm")
298            << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::BOUNDED_RELU, 6.f)).set_name(total_path + "depthwise/Relu6")
299            << ConvolutionLayer(
300                1U, 1U, conv_filt,
301                get_weights_accessor(data_path, total_path + "pointwise_weights.npy", DataLayout::NCHW),
302                std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
303                conv_pad_stride_info)
304            .set_name(total_path + "pointwise/Conv2D")
305            << BatchNormalizationLayer(
306                get_weights_accessor(data_path, total_path + "pointwise_BatchNorm_moving_mean.npy"),
307                get_weights_accessor(data_path, total_path + "pointwise_BatchNorm_moving_variance.npy"),
308                get_weights_accessor(data_path, total_path + "pointwise_BatchNorm_gamma.npy"),
309                get_weights_accessor(data_path, total_path + "pointwise_BatchNorm_beta.npy"),
310                0.001f)
311            .set_name(total_path + "pointwise/BatchNorm")
312            << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::BOUNDED_RELU, 6.f)).set_name(total_path + "pointwise/Relu6");
313 
314         return ConcatLayer(std::move(sg));
315     }
316 
get_dwsc_node_qasymm(const std::string & data_path,std::string && param_path,const unsigned int conv_filt,PadStrideInfo dwc_pad_stride_info,PadStrideInfo conv_pad_stride_info,QuantizationInfo depth_weights_quant_info,QuantizationInfo point_weights_quant_info)317     ConcatLayer get_dwsc_node_qasymm(const std::string &data_path, std::string &&param_path,
318                                      const unsigned int conv_filt,
319                                      PadStrideInfo dwc_pad_stride_info, PadStrideInfo conv_pad_stride_info,
320                                      QuantizationInfo depth_weights_quant_info, QuantizationInfo point_weights_quant_info)
321     {
322         std::string total_path = param_path + "_";
323         SubStream   sg(graph);
324 
325         sg << DepthwiseConvolutionLayer(
326                3U, 3U,
327                get_weights_accessor(data_path, total_path + "depthwise_weights.npy"),
328                get_weights_accessor(data_path, total_path + "depthwise_bias.npy"),
329                dwc_pad_stride_info, 1, std::move(depth_weights_quant_info))
330            .set_name(total_path + "depthwise/depthwise")
331            << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LU_BOUNDED_RELU, 6.f)).set_name(total_path + "depthwise/Relu6")
332            << ConvolutionLayer(
333                1U, 1U, conv_filt,
334                get_weights_accessor(data_path, total_path + "pointwise_weights.npy"),
335                get_weights_accessor(data_path, total_path + "pointwise_bias.npy"),
336                conv_pad_stride_info, 1, std::move(point_weights_quant_info))
337            .set_name(total_path + "pointwise/Conv2D")
338            << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LU_BOUNDED_RELU, 6.f)).set_name(total_path + "pointwise/Relu6");
339 
340         return ConcatLayer(std::move(sg));
341     }
342 };
343 
344 /** Main program for MobileNetV1
345  *
346  * Model is based on:
347  *      https://arxiv.org/abs/1704.04861
348  *      "MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications"
349  *      Andrew G. Howard, Menglong Zhu, Bo Chen, Dmitry Kalenichenko, Weijun Wang, Tobias Weyand, Marco Andreetto, Hartwig Adam
350  *
351  * Provenance: download.tensorflow.org/models/mobilenet_v1_2018_08_02/mobilenet_v1_1.0_224.tgz
352  *             download.tensorflow.org/models/mobilenet_v1_2018_08_02/mobilenet_v1_0.75_160.tgz
353  *
354  * @note To list all the possible arguments execute the binary appended with the --help option
355  *
356  * @param[in] argc Number of arguments
357  * @param[in] argv Arguments
358  */
main(int argc,char ** argv)359 int main(int argc, char **argv)
360 {
361     return arm_compute::utils::run_example<GraphMobilenetExample>(argc, argv);
362 }
363