Gitiles
Code Review Sign In
review.mlplatform.org / ml / ComputeLibrary / 34df2e130034b79050504704e1214194bd83dad5 / . / examples / graph_resnet12.cpp
blob: 33f29dd5002c32bd7a2fc3f5c2fa76baac17e77f [file] [log] [blame]
Georgios Pinitasf554be72018-12-03 16:02:47 +0000[diff] [blame]1/*
Michele Di Giorgio1df9cca2019-01-17 13:20:32 +0000[diff] [blame]2 * Copyright (c) 2018-2019 ARM Limited.
Georgios Pinitasf554be72018-12-03 16:02:47 +0000[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"
25#include "support/ToolchainSupport.h"
26#include "utils/CommonGraphOptions.h"
27#include "utils/GraphUtils.h"
28#include "utils/Utils.h"
29
30using namespace arm_compute::utils;
31using namespace arm_compute::graph::frontend;
32using namespace arm_compute::graph_utils;
33
34/** Example demonstrating how to implement ResNet12 network using the Compute Library's graph API */
35class GraphResNet12Example : public Example
36{
37public:
38 GraphResNet12Example()
39 : cmd_parser(), common_opts(cmd_parser), model_input_width(nullptr), model_input_height(nullptr), common_params(), graph(0, "ResNet12")
40 {
41 model_input_width = cmd_parser.add_option<SimpleOption<unsigned int>>("image-width", 192);
42 model_input_height = cmd_parser.add_option<SimpleOption<unsigned int>>("image-height", 128);
43
44 // Add model id option
45 model_input_width->set_help("Input image width.");
46 model_input_height->set_help("Input image height.");
47 }
48 GraphResNet12Example(const GraphResNet12Example &) = delete;
49 GraphResNet12Example &operator=(const GraphResNet12Example &) = delete;
50 GraphResNet12Example(GraphResNet12Example &&) = default; // NOLINT
51 GraphResNet12Example &operator=(GraphResNet12Example &&) = default; // NOLINT
52 ~GraphResNet12Example() override = default;
53 bool do_setup(int argc, char **argv) override
54 {
55 // Parse arguments
56 cmd_parser.parse(argc, argv);
Georgios Pinitascd60a5f2019-08-21 17:06:54 +0100[diff] [blame]57 cmd_parser.validate();
Georgios Pinitasf554be72018-12-03 16:02:47 +0000[diff] [blame]58
59 // Consume common parameters
60 common_params = consume_common_graph_parameters(common_opts);
61
62 // Return when help menu is requested
63 if(common_params.help)
64 {
65 cmd_parser.print_help(argv[0]);
66 return false;
67 }
68
69 // Get input image width and height
70 const unsigned int image_width = model_input_width->value();
71 const unsigned int image_height = model_input_height->value();
72
73 // Checks
74 ARM_COMPUTE_EXIT_ON_MSG(arm_compute::is_data_type_quantized_asymmetric(common_params.data_type), "QASYMM8 not supported for this graph");
75
76 // Print parameter values
77 std::cout << common_params << std::endl;
78 std::cout << "Image width: " << image_width << std::endl;
79 std::cout << "Image height: " << image_height << std::endl;
80
81 // Get trainable parameters data path
82 const std::string data_path = common_params.data_path;
83 const std::string model_path = "/cnn_data/resnet12_model/";
84
85 // Create a preprocessor object
86 std::unique_ptr<IPreprocessor> preprocessor = arm_compute::support::cpp14::make_unique<TFPreproccessor>();
87
88 // Create input descriptor
89 const TensorShape tensor_shape = permute_shape(TensorShape(image_width, image_height, 3U, 1U), DataLayout::NCHW, common_params.data_layout);
90 TensorDescriptor input_descriptor = TensorDescriptor(tensor_shape, common_params.data_type).set_layout(common_params.data_layout);
91
92 // Set weights trained layout
93 const DataLayout weights_layout = DataLayout::NCHW;
94
95 graph << common_params.target
96 << common_params.fast_math_hint
97 << InputLayer(input_descriptor, get_input_accessor(common_params, std::move(preprocessor), false /* Do not convert to BGR */))
98 << ConvolutionLayer(
99 9U, 9U, 64U,
100 get_weights_accessor(data_path, "conv1_weights.npy", weights_layout),
101 get_weights_accessor(data_path, "conv1_biases.npy", weights_layout),
102 PadStrideInfo(1, 1, 4, 4))
103 .set_name("conv1/convolution")
104 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name("conv1/Relu");
105
106 add_residual_block(data_path, "block1", weights_layout);
107 add_residual_block(data_path, "block2", weights_layout);
108 add_residual_block(data_path, "block3", weights_layout);
109 add_residual_block(data_path, "block4", weights_layout);
110
111 graph << ConvolutionLayer(
112 3U, 3U, 64U,
113 get_weights_accessor(data_path, "conv10_weights.npy", weights_layout),
114 get_weights_accessor(data_path, "conv10_biases.npy"),
115 PadStrideInfo(1, 1, 1, 1))
116 .set_name("conv10/convolution")
117 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name("conv10/Relu")
118 << ConvolutionLayer(
119 3U, 3U, 64U,
120 get_weights_accessor(data_path, "conv11_weights.npy", weights_layout),
121 get_weights_accessor(data_path, "conv11_biases.npy"),
122 PadStrideInfo(1, 1, 1, 1))
123 .set_name("conv11/convolution")
124 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name("conv11/Relu")
125 << ConvolutionLayer(
126 9U, 9U, 3U,
127 get_weights_accessor(data_path, "conv12_weights.npy", weights_layout),
128 get_weights_accessor(data_path, "conv12_biases.npy"),
129 PadStrideInfo(1, 1, 4, 4))
130 .set_name("conv12/convolution")
131 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::TANH)).set_name("conv12/Tanh")
132 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LINEAR, 0.58f, 0.5f)).set_name("conv12/Linear")
133 << OutputLayer(arm_compute::support::cpp14::make_unique<DummyAccessor>(0));
134
135 // Finalize graph
136 GraphConfig config;
137 config.num_threads = common_params.threads;
138 config.use_tuner = common_params.enable_tuner;
Vidhya Sudhan Loganathan050471e2019-04-25 09:27:24 +0100[diff] [blame]139 config.tuner_mode = common_params.tuner_mode;
Michele Di Giorgio1df9cca2019-01-17 13:20:32 +0000[diff] [blame]140 config.tuner_file = common_params.tuner_file;
141
Georgios Pinitasf554be72018-12-03 16:02:47 +0000[diff] [blame]142 graph.finalize(common_params.target, config);
143
144 return true;
145 }
146
147 void do_run() override
148 {
149 // Run graph
150 graph.run();
151 }
152
153private:
154 CommandLineParser cmd_parser;
155 CommonGraphOptions common_opts;
156 SimpleOption<unsigned int> *model_input_width{ nullptr };
157 SimpleOption<unsigned int> *model_input_height{ nullptr };
158 CommonGraphParams common_params;
159 Stream graph;
160
161 void add_residual_block(const std::string &data_path, const std::string &name, DataLayout weights_layout)
162 {
163 std::stringstream unit_path_ss;
164 unit_path_ss << data_path << name << "_";
165 std::stringstream unit_name_ss;
166 unit_name_ss << name << "/";
167
168 std::string unit_path = unit_path_ss.str();
169 std::string unit_name = unit_name_ss.str();
170
171 SubStream left(graph);
172 SubStream right(graph);
173
174 right << ConvolutionLayer(
175 3U, 3U, 64U,
176 get_weights_accessor(data_path, unit_path + "conv1_weights.npy", weights_layout),
177 get_weights_accessor(data_path, unit_path + "conv1_biases.npy", weights_layout),
178 PadStrideInfo(1, 1, 1, 1))
179 .set_name(unit_name + "conv1/convolution")
180 << BatchNormalizationLayer(
181 get_weights_accessor(data_path, unit_path + "conv1_BatchNorm_moving_mean.npy"),
182 get_weights_accessor(data_path, unit_path + "conv1_BatchNorm_moving_variance.npy"),
183 get_weights_accessor(data_path, unit_path + "conv1_BatchNorm_gamma.npy"),
184 get_weights_accessor(data_path, unit_path + "conv1_BatchNorm_beta.npy"),
185 0.0000100099996416f)
186 .set_name(unit_name + "conv1/BatchNorm")
187 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(unit_name + "conv1/Relu")
188
189 << ConvolutionLayer(
190 3U, 3U, 64U,
191 get_weights_accessor(data_path, unit_path + "conv2_weights.npy", weights_layout),
192 get_weights_accessor(data_path, unit_path + "conv2_biases.npy", weights_layout),
193 PadStrideInfo(1, 1, 1, 1))
194 .set_name(unit_name + "conv2/convolution")
195 << BatchNormalizationLayer(
196 get_weights_accessor(data_path, unit_path + "conv2_BatchNorm_moving_mean.npy"),
197 get_weights_accessor(data_path, unit_path + "conv2_BatchNorm_moving_variance.npy"),
198 get_weights_accessor(data_path, unit_path + "conv2_BatchNorm_gamma.npy"),
199 get_weights_accessor(data_path, unit_path + "conv2_BatchNorm_beta.npy"),
200 0.0000100099996416f)
201 .set_name(unit_name + "conv2/BatchNorm")
202 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(unit_name + "conv2/Relu");
203
204 graph << EltwiseLayer(std::move(left), std::move(right), EltwiseOperation::Add).set_name(unit_name + "add");
205 }
206};
207
208/** Main program for ResNet12
209 *
210 * Model is based on:
211 * https://arxiv.org/pdf/1709.01118.pdf
212 * "WESPE: Weakly Supervised Photo Enhancer for Digital Cameras"
213 * Andrey Ignatov, Nikolay Kobyshev, Kenneth Vanhoey, Radu Timofte, Luc Van Gool
214 *
215 * @note To list all the possible arguments execute the binary appended with the --help option
216 *
217 * @param[in] argc Number of arguments
218 * @param[in] argv Arguments
219 */
220int main(int argc, char **argv)
221{
222 return arm_compute::utils::run_example<GraphResNet12Example>(argc, argv);
223}