Gitiles
Code Review Sign In
review.mlplatform.org / ml / ComputeLibrary / 09849a0e7128731473f37cf6045147db68b1c495 / . / examples / graph_vgg19.cpp
blob: 49ae0fe51c4d00ffc75930e12bedecbe5e1bf0f7 [file] [log] [blame]
Isabella Gottardi9f20bda2017-11-03 17:16:20 +0000[diff] [blame]1/*
2 * Copyright (c) 2017 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 */
Isabella Gottardi9f20bda2017-11-03 17:16:20 +0000[diff] [blame]24#include "arm_compute/graph/Graph.h"
25#include "arm_compute/graph/Nodes.h"
Isabella Gottardi9f20bda2017-11-03 17:16:20 +0000[diff] [blame]26#include "support/ToolchainSupport.h"
27#include "utils/GraphUtils.h"
28#include "utils/Utils.h"
29
30#include <cstdlib>
31
32using namespace arm_compute::graph;
33using namespace arm_compute::graph_utils;
Isabella Gottardi9f20bda2017-11-03 17:16:20 +0000[diff] [blame]34
35/** Example demonstrating how to implement VGG19's network using the Compute Library's graph API
36 *
37 * @param[in] argc Number of arguments
Gian Marcobfa3b522017-12-12 10:08:38 +0000[diff] [blame]38 * @param[in] argv Arguments ( [optional] Target (0 = NEON, 1 = OpenCL), [optional] Path to the weights folder, [optional] image, [optional] labels )
Isabella Gottardi9f20bda2017-11-03 17:16:20 +0000[diff] [blame]39 */
40void main_graph_vgg19(int argc, const char **argv)
41{
42 std::string data_path; /* Path to the trainable data */
43 std::string image; /* Image data */
44 std::string label; /* Label data */
45
46 constexpr float mean_r = 123.68f; /* Mean value to subtract from red channel */
47 constexpr float mean_g = 116.779f; /* Mean value to subtract from green channel */
48 constexpr float mean_b = 103.939f; /* Mean value to subtract from blue channel */
49
Gian Marcobfa3b522017-12-12 10:08:38 +0000[diff] [blame]50 // Set target. 0 (NEON), 1 (OpenCL). By default it is NEON
51 TargetHint target_hint = set_target_hint(argc > 1 ? std::strtol(argv[1], nullptr, 10) : 0);
52 ConvolutionMethodHint convolution_hint = ConvolutionMethodHint::DIRECT;
53
Isabella Gottardi9f20bda2017-11-03 17:16:20 +0000[diff] [blame]54 // Parse arguments
55 if(argc < 2)
56 {
57 // Print help
Gian Marcobfa3b522017-12-12 10:08:38 +0000[diff] [blame]58 std::cout << "Usage: " << argv[0] << " [target] [path_to_data] [image] [labels]\n\n";
Isabella Gottardi9f20bda2017-11-03 17:16:20 +0000[diff] [blame]59 std::cout << "No data folder provided: using random values\n\n";
60 }
61 else if(argc == 2)
62 {
Gian Marcobfa3b522017-12-12 10:08:38 +0000[diff] [blame]63 std::cout << "Usage: " << argv[0] << " " << argv[1] << " [path_to_data] [image] [labels]\n\n";
64 std::cout << "No data folder provided: using random values\n\n";
Isabella Gottardi9f20bda2017-11-03 17:16:20 +0000[diff] [blame]65 }
66 else if(argc == 3)
67 {
Gian Marcobfa3b522017-12-12 10:08:38 +0000[diff] [blame]68 data_path = argv[2];
69 std::cout << "Usage: " << argv[0] << " " << argv[1] << " " << argv[2] << " [image] [labels]\n\n";
70 std::cout << "No image provided: using random values\n\n";
71 }
72 else if(argc == 4)
73 {
74 data_path = argv[2];
75 image = argv[3];
76 std::cout << "Usage: " << argv[0] << " " << argv[1] << " " << argv[2] << " " << argv[3] << " [labels]\n\n";
Isabella Gottardi9f20bda2017-11-03 17:16:20 +0000[diff] [blame]77 std::cout << "No text file with labels provided: skipping output accessor\n\n";
78 }
79 else
80 {
Gian Marcobfa3b522017-12-12 10:08:38 +0000[diff] [blame]81 data_path = argv[2];
82 image = argv[3];
83 label = argv[4];
Isabella Gottardi9f20bda2017-11-03 17:16:20 +0000[diff] [blame]84 }
85
86 Graph graph;
87
Gian Marcobfa3b522017-12-12 10:08:38 +0000[diff] [blame]88 graph << target_hint
89 << convolution_hint
Isabella Gottardi9f20bda2017-11-03 17:16:20 +0000[diff] [blame]90 << Tensor(TensorInfo(TensorShape(224U, 224U, 3U, 1U), 1, DataType::F32),
91 get_input_accessor(image, mean_r, mean_g, mean_b))
92 // Layer 1
93 << ConvolutionLayer(
94 3U, 3U, 64U,
95 get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv1_1_w.npy"),
96 get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv1_1_b.npy"),
97 PadStrideInfo(1, 1, 1, 1))
98 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
99 << ConvolutionLayer(
100 3U, 3U, 64U,
101 get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv1_2_w.npy"),
102 get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv1_2_b.npy"),
103 PadStrideInfo(1, 1, 1, 1))
104 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
105 << PoolingLayer(PoolingLayerInfo(PoolingType::MAX, 2, PadStrideInfo(2, 2, 0, 0)))
106 // Layer 2
107 << ConvolutionLayer(
108 3U, 3U, 128U,
109 get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv2_1_w.npy"),
110 get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv2_1_b.npy"),
111 PadStrideInfo(1, 1, 1, 1))
112 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
113 << ConvolutionLayer(
114 3U, 3U, 128U,
115 get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv2_2_w.npy"),
116 get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv2_2_b.npy"),
117 PadStrideInfo(1, 1, 1, 1))
118 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
119 << PoolingLayer(PoolingLayerInfo(PoolingType::MAX, 2, PadStrideInfo(2, 2, 0, 0)))
120 // Layer 3
121 << ConvolutionLayer(
122 3U, 3U, 256U,
123 get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv3_1_w.npy"),
124 get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv3_1_b.npy"),
125 PadStrideInfo(1, 1, 1, 1))
126 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
127 << ConvolutionLayer(
128 3U, 3U, 256U,
129 get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv3_2_w.npy"),
130 get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv3_2_b.npy"),
131 PadStrideInfo(1, 1, 1, 1))
132 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
133 << ConvolutionLayer(
134 3U, 3U, 256U,
135 get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv3_3_w.npy"),
136 get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv3_3_b.npy"),
137 PadStrideInfo(1, 1, 1, 1))
138 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
139 << ConvolutionLayer(
140 3U, 3U, 256U,
141 get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv3_4_w.npy"),
142 get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv3_4_b.npy"),
143 PadStrideInfo(1, 1, 1, 1))
144 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
145 << PoolingLayer(PoolingLayerInfo(PoolingType::MAX, 2, PadStrideInfo(2, 2, 0, 0)))
146 // Layer 4
147 << ConvolutionLayer(
148 3U, 3U, 512U,
149 get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv4_1_w.npy"),
150 get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv4_1_b.npy"),
151 PadStrideInfo(1, 1, 1, 1))
152 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
153 << ConvolutionLayer(
154 3U, 3U, 512U,
155 get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv4_2_w.npy"),
156 get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv4_2_b.npy"),
157 PadStrideInfo(1, 1, 1, 1))
158 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
159 << ConvolutionLayer(
160 3U, 3U, 512U,
161 get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv4_3_w.npy"),
162 get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv4_3_b.npy"),
163 PadStrideInfo(1, 1, 1, 1))
164 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
165 << ConvolutionLayer(
166 3U, 3U, 512U,
167 get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv4_4_w.npy"),
168 get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv4_4_b.npy"),
169 PadStrideInfo(1, 1, 1, 1))
170 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
171 << PoolingLayer(PoolingLayerInfo(PoolingType::MAX, 2, PadStrideInfo(2, 2, 0, 0)))
172 // Layer 5
173 << ConvolutionLayer(
174 3U, 3U, 512U,
175 get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv5_1_w.npy"),
176 get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv5_1_b.npy"),
177 PadStrideInfo(1, 1, 1, 1))
178 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
179 << ConvolutionLayer(
180 3U, 3U, 512U,
181 get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv5_2_w.npy"),
182 get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv5_2_b.npy"),
183 PadStrideInfo(1, 1, 1, 1))
184 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
185 << ConvolutionLayer(
186 3U, 3U, 512U,
187 get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv5_3_w.npy"),
188 get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv5_3_b.npy"),
189 PadStrideInfo(1, 1, 1, 1))
190 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
191 << ConvolutionLayer(
192 3U, 3U, 512U,
193 get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv5_4_w.npy"),
194 get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv5_4_b.npy"),
195 PadStrideInfo(1, 1, 1, 1))
196 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
197 << PoolingLayer(PoolingLayerInfo(PoolingType::MAX, 2, PadStrideInfo(2, 2, 0, 0)))
198 // Layer 6
199 << FullyConnectedLayer(
200 4096U,
201 get_weights_accessor(data_path, "/cnn_data/vgg19_model/fc6_w.npy"),
202 get_weights_accessor(data_path, "/cnn_data/vgg19_model/fc6_b.npy"))
203 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
204 // Layer 7
205 << FullyConnectedLayer(
206 4096U,
207 get_weights_accessor(data_path, "/cnn_data/vgg19_model/fc7_w.npy"),
208 get_weights_accessor(data_path, "/cnn_data/vgg19_model/fc7_b.npy"))
209 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
210 // Layer 8
211 << FullyConnectedLayer(
212 1000U,
213 get_weights_accessor(data_path, "/cnn_data/vgg19_model/fc8_w.npy"),
214 get_weights_accessor(data_path, "/cnn_data/vgg19_model/fc8_b.npy"))
215 // Softmax
216 << SoftmaxLayer()
217 << Tensor(get_output_accessor(label, 5));
218
219 // Run graph
220 graph.run();
221}
222
223/** Main program for VGG19
224 *
225 * @param[in] argc Number of arguments
Gian Marcobfa3b522017-12-12 10:08:38 +0000[diff] [blame]226 * @param[in] argv Arguments ( [optional] Target (0 = NEON, 1 = OpenCL), [optional] Path to the weights folder, [optional] image, [optional] labels )
Isabella Gottardi9f20bda2017-11-03 17:16:20 +0000[diff] [blame]227 */
228int main(int argc, const char **argv)
229{
230 return arm_compute::utils::run_example(argc, argv, main_graph_vgg19);
231}