Gitiles
Code Review Sign In
gerrit-public.fairphone.software / platform / external / ComputeLibrary / 67c8c91522e5be8156b77f57e63c0253535c902a / . / examples / graph_inception_v3.cpp
blob: a55b34e1ed18904e1c564a11e3c83b36e5facf80 [file] [log] [blame]
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]1/*
2 * Copyright (c) 2017-2018 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/Graph.h"
25#include "arm_compute/graph/Nodes.h"
26#include "arm_compute/graph/SubGraph.h"
27#include "support/ToolchainSupport.h"
28#include "utils/GraphUtils.h"
29#include "utils/Utils.h"
30
31#include <cstdlib>
32#include <tuple>
33
34using namespace arm_compute::utils;
35using namespace arm_compute::graph;
36using namespace arm_compute::graph_utils;
37
38/** Example demonstrating how to implement InceptionV3's network using the Compute Library's graph API
39 *
40 * @param[in] argc Number of arguments
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]41 * @param[in] argv Arguments ( [optional] Target (0 = NEON, 1 = OpenCL, 2 = OpenCL with Tuner), [optional] Path to the weights folder, [optional] image, [optional] labels )
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]42 */
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]43class InceptionV3Example final : public Example
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]44{
45public:
46 void do_setup(int argc, char **argv) override
47 {
48 std::string data_path; /* Path to the trainable data */
49 std::string image; /* Image data */
50 std::string label; /* Label data */
51
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]52 // Create a preprocessor object
53 std::unique_ptr<IPreprocessor> preprocessor = arm_compute::support::cpp14::make_unique<TFPreproccessor>();
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]54
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]55 // Set target. 0 (NEON), 1 (OpenCL), 2 (OpenCL with Tuner). By default it is NEON
56 const int int_target_hint = argc > 1 ? std::strtol(argv[1], nullptr, 10) : 0;
57 TargetHint target_hint = set_target_hint(int_target_hint);
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]58
59 // Parse arguments
60 if(argc < 2)
61 {
62 // Print help
63 std::cout << "Usage: " << argv[0] << " [target] [path_to_data] [image] [labels]\n\n";
64 std::cout << "No data folder provided: using random values\n\n";
65 }
66 else if(argc == 2)
67 {
68 std::cout << "Usage: " << argv[0] << " " << argv[1] << " [path_to_data] [image] [labels]\n\n";
69 std::cout << "No data folder provided: using random values\n\n";
70 }
71 else if(argc == 3)
72 {
73 data_path = argv[2];
74 std::cout << "Usage: " << argv[0] << " " << argv[1] << " " << argv[2] << " [image] [labels]\n\n";
75 std::cout << "No image provided: using random values\n\n";
76 }
77 else if(argc == 4)
78 {
79 data_path = argv[2];
80 image = argv[3];
81 std::cout << "Usage: " << argv[0] << " " << argv[1] << " " << argv[2] << " " << argv[3] << " [labels]\n\n";
82 std::cout << "No text file with labels provided: skipping output accessor\n\n";
83 }
84 else
85 {
86 data_path = argv[2];
87 image = argv[3];
88 label = argv[4];
89 }
90
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]91 graph << target_hint << Tensor(TensorInfo(TensorShape(299U, 299U, 3U, 1U), 1, DataType::F32),
92 get_input_accessor(image, std::move(preprocessor), false))
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]93
94 << ConvolutionLayer(3U, 3U, 32U,
95 get_weights_accessor(data_path, "/cnn_data/inceptionv3_model/Conv2d_1a_3x3_weights.npy"),
96 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(2, 2, 0, 0))
97 << BatchNormalizationLayer(get_weights_accessor(data_path,
98 "/cnn_data/inceptionv3_model/Conv2d_1a_3x3_BatchNorm_moving_mean.npy"),
99 get_weights_accessor(data_path,
100 "/cnn_data/inceptionv3_model/Conv2d_1a_3x3_BatchNorm_moving_variance.npy"),
101 get_random_accessor(1.f, 1.f), get_weights_accessor(data_path,
102 "/cnn_data/inceptionv3_model/Conv2d_1a_3x3_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]103 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]104
105 << ConvolutionLayer(3U, 3U, 32U,
106 get_weights_accessor(data_path, "/cnn_data/inceptionv3_model/Conv2d_2a_3x3_weights.npy"),
107 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 0, 0))
108 << BatchNormalizationLayer(get_weights_accessor(data_path,
109 "/cnn_data/inceptionv3_model/Conv2d_2a_3x3_BatchNorm_moving_mean.npy"),
110 get_weights_accessor(data_path,
111 "/cnn_data/inceptionv3_model/Conv2d_2a_3x3_BatchNorm_moving_variance.npy"),
112 get_random_accessor(1.f, 1.f), get_weights_accessor(data_path,
113 "/cnn_data/inceptionv3_model/Conv2d_2a_3x3_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]114 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]115
116 << ConvolutionLayer(3U, 3U, 64U,
117 get_weights_accessor(data_path, "/cnn_data/inceptionv3_model/Conv2d_2b_3x3_weights.npy"),
118 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 1, 1))
119 << BatchNormalizationLayer(get_weights_accessor(data_path,
120 "/cnn_data/inceptionv3_model/Conv2d_2b_3x3_BatchNorm_moving_mean.npy"),
121 get_weights_accessor(data_path,
122 "/cnn_data/inceptionv3_model/Conv2d_2b_3x3_BatchNorm_moving_variance.npy"),
123 get_random_accessor(1.f, 1.f), get_weights_accessor(data_path,
124 "/cnn_data/inceptionv3_model/Conv2d_2b_3x3_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]125 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]126
127 << PoolingLayer(PoolingLayerInfo(PoolingType::MAX, 3, PadStrideInfo(2, 2, 0, 0, DimensionRoundingType::CEIL)))
128
129 << ConvolutionLayer(1U, 1U, 80U,
130 get_weights_accessor(data_path, "/cnn_data/inceptionv3_model/Conv2d_3b_1x1_weights.npy"),
131 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 0, 0))
132 << BatchNormalizationLayer(get_weights_accessor(data_path,
133 "/cnn_data/inceptionv3_model/Conv2d_3b_1x1_BatchNorm_moving_mean.npy"),
134 get_weights_accessor(data_path,
135 "/cnn_data/inceptionv3_model/Conv2d_3b_1x1_BatchNorm_moving_variance.npy"),
136 get_random_accessor(1.f, 1.f), get_weights_accessor(data_path,
137 "/cnn_data/inceptionv3_model/Conv2d_3b_1x1_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]138 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]139
140 << ConvolutionLayer(3U, 3U, 192U,
141 get_weights_accessor(data_path, "/cnn_data/inceptionv3_model/Conv2d_4a_3x3_weights.npy"),
142 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 0, 0))
143 << BatchNormalizationLayer(get_weights_accessor(data_path,
144 "/cnn_data/inceptionv3_model/Conv2d_4a_3x3_BatchNorm_moving_mean.npy"),
145 get_weights_accessor(data_path,
146 "/cnn_data/inceptionv3_model/Conv2d_4a_3x3_BatchNorm_moving_variance.npy"),
147 get_random_accessor(1.f, 1.f), get_weights_accessor(data_path,
148 "/cnn_data/inceptionv3_model/Conv2d_4a_3x3_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]149 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]150
151 << PoolingLayer(PoolingLayerInfo(PoolingType::MAX, 3, PadStrideInfo(2, 2, 0, 0, DimensionRoundingType::CEIL)))
152
153 << get_inception_node_A(data_path, "Mixed_5b", 64U, std::make_tuple(48U, 64U), std::make_tuple(64U, 96U, 96U),
154 32U)
155 << get_inception_node_A(data_path, "Mixed_5c", 64U, std::make_tuple(48U, 64U), std::make_tuple(64U, 96U, 96U),
156 64U, true)
157 << get_inception_node_A(data_path, "Mixed_5d", 64U, std::make_tuple(48U, 64U), std::make_tuple(64U, 96U, 96U),
158 64U)
159
160 << get_inception_node_B(data_path, "Mixed_6a", 384U, std::make_tuple(64U, 96U, 96U))
161
162 << get_inception_node_C(data_path, "Mixed_6b", 192U, std::make_tuple(128U, 128U, 192U),
163 std::make_tuple(128U, 128U, 128U, 128U, 192U), 192U)
164 << get_inception_node_C(data_path, "Mixed_6c", 192U, std::make_tuple(160U, 160U, 192U),
165 std::make_tuple(160U, 160U, 160U, 160U, 192U), 192U)
166 << get_inception_node_C(data_path, "Mixed_6d", 192U, std::make_tuple(160U, 160U, 192U),
167 std::make_tuple(160U, 160U, 160U, 160U, 192U), 192U)
168 << get_inception_node_C(data_path, "Mixed_6e", 192U, std::make_tuple(192U, 192U, 192U),
169 std::make_tuple(192U, 192U, 192U, 192U, 192U), 192U)
170
171 << get_inception_node_D(data_path, "Mixed_7a", std::make_tuple(192U, 320U),
172 std::make_tuple(192U, 192U, 192U, 192U))
173
174 << get_inception_node_E(data_path, "Mixed_7b", 320U, std::make_tuple(384U, 384U, 384U),
175 std::make_tuple(448U, 384U, 384U, 384U), 192U)
176 << get_inception_node_E(data_path, "Mixed_7c", 320U, std::make_tuple(384U, 384U, 384U),
177 std::make_tuple(448U, 384U, 384U, 384U), 192U, true)
178
179 << PoolingLayer(PoolingLayerInfo(PoolingType::AVG, 8, PadStrideInfo(1, 1, 0, 0, DimensionRoundingType::CEIL)))
180 << ConvolutionLayer(1U, 1U, 1001U, get_weights_accessor(data_path,
181 "/cnn_data/inceptionv3_model/Logits_Conv2d_1c_1x1_weights.npy"),
182 get_weights_accessor(data_path,
183 "/cnn_data/inceptionv3_model/Logits_Conv2d_1c_1x1_biases.npy"),
184 PadStrideInfo(1, 1, 0, 0))
185 << ReshapeLayer(TensorShape(1001U)) << SoftmaxLayer()
186 << Tensor(get_output_accessor(label, 5));
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]187
188 // In order to enable the OpenCL tuner, graph_init() has to be called only when all nodes have been instantiated
189 graph.graph_init(int_target_hint == 2);
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]190 }
191
192 void do_run() override
193 {
194 graph.run();
195 }
196
197private:
198 Graph graph{};
199
200private:
201 BranchLayer get_inception_node_A(const std::string &data_path, std::string &&param_path,
202 unsigned int a_filt,
203 std::tuple<unsigned int, unsigned int> b_filters,
204 std::tuple<unsigned int, unsigned int, unsigned int> c_filters,
205 unsigned int d_filt,
206 bool is_name_different = false)
207 {
208 std::string total_path = "/cnn_data/inceptionv3_model/" + param_path + "_";
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]209
210 // This is due to a naming issue in the tf model
211 std::string conv_id0 = "_0a_";
212 std::string conv_id1 = "2d_0b_";
213 if(is_name_different)
214 {
215 conv_id0 = "_0b_";
216 conv_id1 = "_1_0c_";
217 }
218
219 SubGraph i_a;
220 i_a << ConvolutionLayer(
221 1U, 1U, a_filt,
222 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_weights.npy"),
223 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
224 PadStrideInfo(1, 1, 0, 0))
225 << BatchNormalizationLayer(
226 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
227 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
228 get_random_accessor(1.f, 1.f),
229 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]230 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU));
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]231
232 SubGraph i_b;
233 i_b << ConvolutionLayer(
234 1U, 1U, std::get<0>(b_filters),
235 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d" + conv_id0 + "1x1_weights.npy"),
236 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
237 PadStrideInfo(1, 1, 0, 0))
238 << BatchNormalizationLayer(
239 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d" + conv_id0 + "1x1_BatchNorm_moving_mean.npy"),
240 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d" + conv_id0 + "1x1_BatchNorm_moving_variance.npy"),
241 get_random_accessor(1.f, 1.f),
242 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d" + conv_id0 + "1x1_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]243 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]244 << ConvolutionLayer(
245 5U, 5U, std::get<1>(b_filters),
246 get_weights_accessor(data_path, total_path + "Branch_1_Conv" + conv_id1 + "5x5_weights.npy"),
247 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
248 PadStrideInfo(1, 1, 2, 2))
249 << BatchNormalizationLayer(
250 get_weights_accessor(data_path, total_path + "Branch_1_Conv" + conv_id1 + "5x5_BatchNorm_moving_mean.npy"),
251 get_weights_accessor(data_path, total_path + "Branch_1_Conv" + conv_id1 + "5x5_BatchNorm_moving_variance.npy"),
252 get_random_accessor(1.f, 1.f),
253 get_weights_accessor(data_path, total_path + "Branch_1_Conv" + conv_id1 + "5x5_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]254 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU));
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]255
256 SubGraph i_c;
257 i_c << ConvolutionLayer(
258 1U, 1U, std::get<0>(c_filters),
259 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_weights.npy"),
260 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
261 PadStrideInfo(1, 1, 0, 0))
262 << BatchNormalizationLayer(
263 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
264 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
265 get_random_accessor(1.f, 1.f),
266 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]267 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]268 << ConvolutionLayer(
269 3U, 3U, std::get<1>(c_filters),
270 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_3x3_weights.npy"),
271 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
272 PadStrideInfo(1, 1, 1, 1))
273 << BatchNormalizationLayer(
274 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_3x3_BatchNorm_moving_mean.npy"),
275 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_3x3_BatchNorm_moving_variance.npy"),
276 get_random_accessor(1.f, 1.f),
277 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_3x3_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]278 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]279 << ConvolutionLayer(
280 3U, 3U, std::get<2>(c_filters),
281 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_3x3_weights.npy"),
282 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
283 PadStrideInfo(1, 1, 1, 1))
284 << BatchNormalizationLayer(
285 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_3x3_BatchNorm_moving_mean.npy"),
286 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_3x3_BatchNorm_moving_variance.npy"),
287 get_random_accessor(1.f, 1.f),
288 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_3x3_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]289 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU));
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]290
291 SubGraph i_d;
292 i_d << PoolingLayer(PoolingLayerInfo(PoolingType::AVG, 3, PadStrideInfo(1, 1, 1, 1, DimensionRoundingType::CEIL), true))
293 << ConvolutionLayer(
294 1U, 1U, d_filt,
295 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_weights.npy"),
296 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
297 PadStrideInfo(1, 1, 0, 0))
298 << BatchNormalizationLayer(
299 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_BatchNorm_moving_mean.npy"),
300 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_BatchNorm_moving_variance.npy"),
301 get_random_accessor(1.f, 1.f),
302 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]303 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU));
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]304
305 return BranchLayer(BranchMergeMethod::DEPTH_CONCATENATE, std::move(i_a), std::move(i_b), std::move(i_c), std::move(i_d));
306 }
307
308 BranchLayer get_inception_node_B(const std::string &data_path, std::string &&param_path,
309 unsigned int a_filt,
310 std::tuple<unsigned int, unsigned int, unsigned int> b_filters)
311 {
312 std::string total_path = "/cnn_data/inceptionv3_model/" + param_path + "_";
313 SubGraph i_a;
314 i_a << ConvolutionLayer(
315 3U, 3U, a_filt,
316 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_1a_1x1_weights.npy"),
317 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
318 PadStrideInfo(2, 2, 0, 0))
319 << BatchNormalizationLayer(
320 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_1a_1x1_BatchNorm_moving_mean.npy"),
321 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_1a_1x1_BatchNorm_moving_variance.npy"),
322 get_random_accessor(1.f, 1.f),
323 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_1a_1x1_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]324 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU));
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]325
326 SubGraph i_b;
327 i_b << ConvolutionLayer(
328 1U, 1U, std::get<0>(b_filters),
329 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_weights.npy"),
330 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
331 PadStrideInfo(1, 1, 0, 0))
332 << BatchNormalizationLayer(
333 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
334 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
335 get_random_accessor(1.f, 1.f),
336 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]337 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]338 << ConvolutionLayer(
339 3U, 3U, std::get<1>(b_filters),
340 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_3x3_weights.npy"),
341 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
342 PadStrideInfo(1, 1, 1, 1))
343 << BatchNormalizationLayer(
344 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_3x3_BatchNorm_moving_mean.npy"),
345 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_3x3_BatchNorm_moving_variance.npy"),
346 get_random_accessor(1.f, 1.f),
347 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_3x3_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]348 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]349 << ConvolutionLayer(
350 3U, 3U, std::get<2>(b_filters),
351 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_1a_1x1_weights.npy"),
352 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
353 PadStrideInfo(2, 2, 0, 0))
354 << BatchNormalizationLayer(
355 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_1a_1x1_BatchNorm_moving_mean.npy"),
356 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_1a_1x1_BatchNorm_moving_variance.npy"),
357 get_random_accessor(1.f, 1.f),
358 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_1a_1x1_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]359 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU));
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]360
361 SubGraph i_c;
362 i_c << PoolingLayer(PoolingLayerInfo(PoolingType::MAX, 3, PadStrideInfo(2, 2, 0, 0, DimensionRoundingType::CEIL)))
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]363 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LINEAR, 1.f, 0.f));
364
365 return BranchLayer(BranchMergeMethod::DEPTH_CONCATENATE, std::move(i_a), std::move(i_b), std::move(i_c));
366 }
367
368 BranchLayer get_inception_node_C(const std::string &data_path, std::string &&param_path,
369 unsigned int a_filt,
370 std::tuple<unsigned int, unsigned int, unsigned int> b_filters,
371 std::tuple<unsigned int, unsigned int, unsigned int, unsigned int, unsigned int> c_filters,
372 unsigned int d_filt)
373 {
374 std::string total_path = "/cnn_data/inceptionv3_model/" + param_path + "_";
375 SubGraph i_a;
376 i_a << ConvolutionLayer(
377 1U, 1U, a_filt,
378 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_weights.npy"),
379 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
380 PadStrideInfo(1, 1, 0, 0))
381 << BatchNormalizationLayer(
382 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
383 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
384 get_random_accessor(1.f, 1.f),
385 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]386 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU));
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]387
388 SubGraph i_b;
389 i_b << ConvolutionLayer(
390 1U, 1U, std::get<0>(b_filters),
391 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_weights.npy"),
392 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
393 PadStrideInfo(1, 1, 0, 0))
394 << BatchNormalizationLayer(
395 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
396 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
397 get_random_accessor(1.f, 1.f),
398 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]399 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]400 << ConvolutionLayer(
401 7U, 1U, std::get<1>(b_filters),
402 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x7_weights.npy"),
403 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
404 PadStrideInfo(1, 1, 3, 0))
405 << BatchNormalizationLayer(
406 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x7_BatchNorm_moving_mean.npy"),
407 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x7_BatchNorm_moving_variance.npy"),
408 get_random_accessor(1.f, 1.f),
409 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x7_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]410 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]411 << ConvolutionLayer(
412 1U, 7U, std::get<2>(b_filters),
413 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0c_7x1_weights.npy"),
414 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
415 PadStrideInfo(1, 1, 0, 3))
416 << BatchNormalizationLayer(
417 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0c_7x1_BatchNorm_moving_mean.npy"),
418 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0c_7x1_BatchNorm_moving_variance.npy"),
419 get_random_accessor(1.f, 1.f),
420 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0c_7x1_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]421 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU));
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]422
423 SubGraph i_c;
424 i_c << ConvolutionLayer(
425 1U, 1U, std::get<0>(c_filters),
426 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_weights.npy"),
427 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
428 PadStrideInfo(1, 1, 0, 0))
429 << BatchNormalizationLayer(
430 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
431 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
432 get_random_accessor(1.f, 1.f),
433 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]434 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]435 << ConvolutionLayer(
436 1U, 7U, std::get<1>(c_filters),
437 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_7x1_weights.npy"),
438 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
439 PadStrideInfo(1, 1, 0, 3))
440 << BatchNormalizationLayer(
441 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_7x1_BatchNorm_moving_mean.npy"),
442 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_7x1_BatchNorm_moving_variance.npy"),
443 get_random_accessor(1.f, 1.f),
444 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_7x1_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]445 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]446 << ConvolutionLayer(
447 7U, 1U, std::get<2>(c_filters),
448 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_1x7_weights.npy"),
449 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
450 PadStrideInfo(1, 1, 3, 0))
451 << BatchNormalizationLayer(
452 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_1x7_BatchNorm_moving_mean.npy"),
453 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_1x7_BatchNorm_moving_variance.npy"),
454 get_random_accessor(1.f, 1.f),
455 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_1x7_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]456 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]457 << ConvolutionLayer(
458 1U, 7U, std::get<3>(c_filters),
459 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0d_7x1_weights.npy"),
460 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
461 PadStrideInfo(1, 1, 0, 3))
462 << BatchNormalizationLayer(
463 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0d_7x1_BatchNorm_moving_mean.npy"),
464 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0d_7x1_BatchNorm_moving_variance.npy"),
465 get_random_accessor(1.f, 1.f),
466 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0d_7x1_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]467 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]468 << ConvolutionLayer(
469 7U, 1U, std::get<4>(c_filters),
470 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0e_1x7_weights.npy"),
471 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
472 PadStrideInfo(1, 1, 3, 0))
473 << BatchNormalizationLayer(
474 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0e_1x7_BatchNorm_moving_mean.npy"),
475 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0e_1x7_BatchNorm_moving_variance.npy"),
476 get_random_accessor(1.f, 1.f),
477 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0e_1x7_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]478 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU));
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]479
480 SubGraph i_d;
481 i_d << PoolingLayer(PoolingLayerInfo(PoolingType::AVG, 3, PadStrideInfo(1, 1, 1, 1, DimensionRoundingType::CEIL), true))
482 << ConvolutionLayer(
483 1U, 1U, d_filt,
484 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_weights.npy"),
485 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
486 PadStrideInfo(1, 1, 0, 0))
487 << BatchNormalizationLayer(
488 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_BatchNorm_moving_mean.npy"),
489 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_BatchNorm_moving_variance.npy"),
490 get_random_accessor(1.f, 1.f),
491 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]492 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU));
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]493
494 return BranchLayer(BranchMergeMethod::DEPTH_CONCATENATE, std::move(i_a), std::move(i_b), std::move(i_c), std::move(i_d));
495 }
496
497 BranchLayer get_inception_node_D(const std::string &data_path, std::string &&param_path,
498 std::tuple<unsigned int, unsigned int> a_filters,
499 std::tuple<unsigned int, unsigned int, unsigned int, unsigned int> b_filters)
500 {
501 std::string total_path = "/cnn_data/inceptionv3_model/" + param_path + "_";
502 SubGraph i_a;
503 i_a << ConvolutionLayer(
504 1U, 1U, std::get<0>(a_filters),
505 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_weights.npy"),
506 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
507 PadStrideInfo(1, 1, 0, 0))
508 << BatchNormalizationLayer(
509 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
510 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
511 get_random_accessor(1.f, 1.f),
512 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]513 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]514 << ConvolutionLayer(
515 3U, 3U, std::get<1>(a_filters),
516 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_1a_3x3_weights.npy"),
517 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
518 PadStrideInfo(2, 2, 0, 0))
519 << BatchNormalizationLayer(
520 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_1a_3x3_BatchNorm_moving_mean.npy"),
521 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_1a_3x3_BatchNorm_moving_variance.npy"),
522 get_random_accessor(1.f, 1.f),
523 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_1a_3x3_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]524 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU));
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]525
526 SubGraph i_b;
527 i_b << ConvolutionLayer(
528 1U, 1U, std::get<0>(b_filters),
529 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_weights.npy"),
530 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
531 PadStrideInfo(1, 1, 0, 0))
532 << BatchNormalizationLayer(
533 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
534 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
535 get_random_accessor(1.f, 1.f),
536 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]537 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]538 << ConvolutionLayer(
539 7U, 1U, std::get<1>(b_filters),
540 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x7_weights.npy"),
541 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
542 PadStrideInfo(1, 1, 3, 0))
543 << BatchNormalizationLayer(
544 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x7_BatchNorm_moving_mean.npy"),
545 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x7_BatchNorm_moving_variance.npy"),
546 get_random_accessor(1.f, 1.f),
547 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x7_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]548 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]549 << ConvolutionLayer(
550 1U, 7U, std::get<2>(b_filters),
551 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0c_7x1_weights.npy"),
552 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
553 PadStrideInfo(1, 1, 0, 3))
554 << BatchNormalizationLayer(
555 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0c_7x1_BatchNorm_moving_mean.npy"),
556 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0c_7x1_BatchNorm_moving_variance.npy"),
557 get_random_accessor(1.f, 1.f),
558 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0c_7x1_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]559 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]560 << ConvolutionLayer(
561 3U, 3U, std::get<3>(b_filters),
562 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_1a_3x3_weights.npy"),
563 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
564 PadStrideInfo(2, 2, 0, 0))
565 << BatchNormalizationLayer(
566 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_1a_3x3_BatchNorm_moving_mean.npy"),
567 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_1a_3x3_BatchNorm_moving_variance.npy"),
568 get_random_accessor(1.f, 1.f),
569 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_1a_3x3_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]570 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU));
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]571
572 SubGraph i_c;
573 i_c << PoolingLayer(PoolingLayerInfo(PoolingType::MAX, 3, PadStrideInfo(2, 2, 0, 0, DimensionRoundingType::CEIL)))
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]574 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LINEAR, 1.f, 0.f));
575
576 return BranchLayer(BranchMergeMethod::DEPTH_CONCATENATE, std::move(i_a), std::move(i_b), std::move(i_c));
577 }
578
579 BranchLayer get_inception_node_E(const std::string &data_path, std::string &&param_path,
580 unsigned int a_filt,
581 std::tuple<unsigned int, unsigned int, unsigned int> b_filters,
582 std::tuple<unsigned int, unsigned int, unsigned int, unsigned int> c_filters,
583 unsigned int d_filt,
584 bool is_name_different = false)
585 {
586 // This is due to a naming issue in the tf model
587 std::string conv_id = "_0b_";
588 if(is_name_different)
589 {
590 conv_id = "_0c_";
591 }
592
593 std::string total_path = "/cnn_data/inceptionv3_model/" + param_path + "_";
594 SubGraph i_a;
595 i_a << ConvolutionLayer(
596 1U, 1U, a_filt,
597 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_weights.npy"),
598 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
599 PadStrideInfo(1, 1, 0, 0))
600 << BatchNormalizationLayer(
601 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
602 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
603 get_random_accessor(1.f, 1.f),
604 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]605 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU));
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]606
607 SubGraph i_b1;
608 i_b1 << ConvolutionLayer(
609 3U, 1U, std::get<1>(b_filters),
610 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x3_weights.npy"),
611 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
612 PadStrideInfo(1, 1, 1, 0))
613 << BatchNormalizationLayer(
614 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x3_BatchNorm_moving_mean.npy"),
615 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x3_BatchNorm_moving_variance.npy"),
616 get_random_accessor(1.f, 1.f),
617 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x3_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]618 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU));
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]619
620 SubGraph i_b2;
621 i_b2 << ConvolutionLayer(
622 1U, 3U, std::get<2>(b_filters),
623 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d" + conv_id + "3x1_weights.npy"),
624 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
625 PadStrideInfo(1, 1, 0, 1))
626 << BatchNormalizationLayer(
627 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d" + conv_id + "3x1_BatchNorm_moving_mean.npy"),
628 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d" + conv_id + "3x1_BatchNorm_moving_variance.npy"),
629 get_random_accessor(1.f, 1.f),
630 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d" + conv_id + "3x1_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]631 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU));
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]632
633 SubGraph i_b;
634 i_b << ConvolutionLayer(
635 1U, 1U, std::get<0>(b_filters),
636 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_weights.npy"),
637 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
638 PadStrideInfo(1, 1, 0, 0))
639 << BatchNormalizationLayer(
640 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
641 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
642 get_random_accessor(1.f, 1.f),
643 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]644 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]645 << BranchLayer(BranchMergeMethod::DEPTH_CONCATENATE, std::move(i_b1), std::move(i_b2));
646
647 SubGraph i_c1;
648 i_c1 << ConvolutionLayer(
649 3U, 1U, std::get<2>(c_filters),
650 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_1x3_weights.npy"),
651 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
652 PadStrideInfo(1, 1, 1, 0))
653 << BatchNormalizationLayer(
654 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_1x3_BatchNorm_moving_mean.npy"),
655 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_1x3_BatchNorm_moving_variance.npy"),
656 get_random_accessor(1.f, 1.f),
657 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_1x3_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]658 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU));
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]659
660 SubGraph i_c2;
661 i_c2 << ConvolutionLayer(
662 1U, 3U, std::get<3>(c_filters),
663 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0d_3x1_weights.npy"),
664 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
665 PadStrideInfo(1, 1, 0, 1))
666 << BatchNormalizationLayer(
667 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0d_3x1_BatchNorm_moving_mean.npy"),
668 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0d_3x1_BatchNorm_moving_variance.npy"),
669 get_random_accessor(1.f, 1.f),
670 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0d_3x1_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]671 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU));
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]672
673 SubGraph i_c;
674 i_c << ConvolutionLayer(
675 1U, 1U, std::get<0>(c_filters),
676 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_weights.npy"),
677 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
678 PadStrideInfo(1, 1, 0, 0))
679 << BatchNormalizationLayer(
680 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
681 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
682 get_random_accessor(1.f, 1.f),
683 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]684 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]685 << ConvolutionLayer(
686 3U, 3U, std::get<1>(c_filters),
687 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_3x3_weights.npy"),
688 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
689 PadStrideInfo(1, 1, 1, 1))
690 << BatchNormalizationLayer(
691 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_3x3_BatchNorm_moving_mean.npy"),
692 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_3x3_BatchNorm_moving_variance.npy"),
693 get_random_accessor(1.f, 1.f),
694 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_3x3_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]695 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]696 << BranchLayer(BranchMergeMethod::DEPTH_CONCATENATE, std::move(i_c1), std::move(i_c2));
697
698 SubGraph i_d;
699 i_d << PoolingLayer(PoolingLayerInfo(PoolingType::AVG, 3, PadStrideInfo(1, 1, 1, 1, DimensionRoundingType::CEIL), true))
700 << ConvolutionLayer(
701 1U, 1U, d_filt,
702 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_weights.npy"),
703 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
704 PadStrideInfo(1, 1, 0, 0))
705 << BatchNormalizationLayer(
706 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_BatchNorm_moving_mean.npy"),
707 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_BatchNorm_moving_variance.npy"),
708 get_random_accessor(1.f, 1.f),
709 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_BatchNorm_beta.npy"),
Anthony Barbier06ea0482018-02-22 15:45:35 +0000[diff] [blame]710 0.001f, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU));
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]711
712 return BranchLayer(BranchMergeMethod::DEPTH_CONCATENATE, std::move(i_a), std::move(i_b), std::move(i_c), std::move(i_d));
713 }
714};
715
716/** Main program for Inception V3
717 *
718 * @param[in] argc Number of arguments
Jenkinsc3f34a42018-03-02 12:38:09 +0000[diff] [blame]719 * @param[in] argv Arguments ( [optional] Target (0 = NEON, 1 = OpenCL, 2 = OpenCL with Tuner), [optional] Path to the weights folder, [optional] image, [optional] labels )
Anthony Barbierf45d5a92018-01-24 16:23:15 +0000[diff] [blame]720 */
721int main(int argc, char **argv)
722{
723 return arm_compute::utils::run_example<InceptionV3Example>(argc, argv);
724}