1 /*
2 * Copyright (c) 2018-2021 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 #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::utils;
31 using namespace arm_compute::graph::frontend;
32 using namespace arm_compute::graph_utils;
33
34 /** Example demonstrating how to implement SRCNN 9-5-5 network using the Compute Library's graph API */
35 class GraphSRCNN955Example : public Example
36 {
37 public:
GraphSRCNN955Example()38 GraphSRCNN955Example()
39 : cmd_parser(), common_opts(cmd_parser), model_input_width(nullptr), model_input_height(nullptr), common_params(), graph(0, "SRCNN955")
40 {
41 model_input_width = cmd_parser.add_option<SimpleOption<unsigned int>>("image-width", 300);
42 model_input_height = cmd_parser.add_option<SimpleOption<unsigned int>>("image-height", 300);
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 GraphSRCNN955Example(const GraphSRCNN955Example &) = delete;
49 GraphSRCNN955Example &operator=(const GraphSRCNN955Example &) = delete;
50 ~GraphSRCNN955Example() override = default;
do_setup(int argc,char ** argv)51 bool do_setup(int argc, char **argv) override
52 {
53 // Parse arguments
54 cmd_parser.parse(argc, argv);
55 cmd_parser.validate();
56
57 // Consume common parameters
58 common_params = consume_common_graph_parameters(common_opts);
59
60 // Return when help menu is requested
61 if(common_params.help)
62 {
63 cmd_parser.print_help(argv[0]);
64 return false;
65 }
66
67 // Get input image width and height
68 const unsigned int image_width = model_input_width->value();
69 const unsigned int image_height = model_input_height->value();
70
71 // Print parameter values
72 std::cout << common_params << std::endl;
73 std::cout << "Image width: " << image_width << std::endl;
74 std::cout << "Image height: " << image_height << std::endl;
75
76 // Get trainable parameters data path
77 const std::string data_path = common_params.data_path;
78 const std::string model_path = "/cnn_data/srcnn955_model/";
79
80 // Create a preprocessor object
81 std::unique_ptr<IPreprocessor> preprocessor = std::make_unique<TFPreproccessor>();
82
83 // Create input descriptor
84 const TensorShape tensor_shape = permute_shape(TensorShape(image_width, image_height, 3U, common_params.batches), DataLayout::NCHW, common_params.data_layout);
85 TensorDescriptor input_descriptor = TensorDescriptor(tensor_shape, common_params.data_type).set_layout(common_params.data_layout);
86
87 // Set weights trained layout
88 const DataLayout weights_layout = DataLayout::NCHW;
89
90 graph << common_params.target
91 << common_params.fast_math_hint
92 << InputLayer(input_descriptor, get_input_accessor(common_params, std::move(preprocessor), false /* Do not convert to BGR */))
93 << ConvolutionLayer(
94 9U, 9U, 64U,
95 get_weights_accessor(data_path, "conv1_weights.npy", weights_layout),
96 get_weights_accessor(data_path, "conv1_biases.npy"),
97 PadStrideInfo(1, 1, 4, 4))
98 .set_name("conv1/convolution")
99 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name("conv1/Relu")
100 << ConvolutionLayer(
101 5U, 5U, 32U,
102 get_weights_accessor(data_path, "conv2_weights.npy", weights_layout),
103 get_weights_accessor(data_path, "conv2_biases.npy"),
104 PadStrideInfo(1, 1, 2, 2))
105 .set_name("conv2/convolution")
106 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name("conv2/Relu")
107 << ConvolutionLayer(
108 5U, 5U, 3U,
109 get_weights_accessor(data_path, "conv3_weights.npy", weights_layout),
110 get_weights_accessor(data_path, "conv3_biases.npy"),
111 PadStrideInfo(1, 1, 2, 2))
112 .set_name("conv3/convolution")
113 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name("conv3/Relu")
114 << OutputLayer(std::make_unique<DummyAccessor>(0));
115
116 // Finalize graph
117 GraphConfig config;
118 config.num_threads = common_params.threads;
119 config.use_tuner = common_params.enable_tuner;
120 config.tuner_mode = common_params.tuner_mode;
121 config.tuner_file = common_params.tuner_file;
122 config.mlgo_file = common_params.mlgo_file;
123 config.use_synthetic_type = arm_compute::is_data_type_quantized(common_params.data_type);
124 config.synthetic_type = common_params.data_type;
125
126 graph.finalize(common_params.target, config);
127
128 return true;
129 }
130
do_run()131 void do_run() override
132 {
133 // Run graph
134 graph.run();
135 }
136
137 private:
138 CommandLineParser cmd_parser;
139 CommonGraphOptions common_opts;
140 SimpleOption<unsigned int> *model_input_width{ nullptr };
141 SimpleOption<unsigned int> *model_input_height{ nullptr };
142 CommonGraphParams common_params;
143 Stream graph;
144 };
145
146 /** Main program for SRCNN 9-5-5
147 *
148 * Model is based on:
149 * http://mmlab.ie.cuhk.edu.hk/projects/SRCNN.html
150 * "Image Super-Resolution Using Deep Convolutional Networks"
151 * Chao Dong, Chen Change Loy, Kaiming He, Xiaoou Tang
152 *
153 * @note To list all the possible arguments execute the binary appended with the --help option
154 *
155 * @param[in] argc Number of arguments
156 * @param[in] argv Arguments
157 */
main(int argc,char ** argv)158 int main(int argc, char **argv)
159 {
160 return arm_compute::utils::run_example<GraphSRCNN955Example>(argc, argv);
161 }
162