Gitiles
Code Review Sign In
review.mlplatform.org / ml / ComputeLibrary / b0b37177f190a261b338cca53b6c6136eea14ba1 / . / examples / graph_inception_v3.cpp
blob: 67f4e3cacf2e353277b67ca01248e3dc754e0b5a [file] [log] [blame]
Georgios Pinitas652bde52018-01-10 15:33:28 +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 */
Georgios Pinitasd9eb2752018-04-03 13:44:29 +0100[diff] [blame]24#include "arm_compute/graph.h"
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]25#include "support/ToolchainSupport.h"
Georgios Pinitas12be7ab2018-07-03 12:06:23 +0100[diff] [blame]26#include "utils/CommonGraphOptions.h"
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]27#include "utils/GraphUtils.h"
28#include "utils/Utils.h"
29
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]30using namespace arm_compute::utils;
Georgios Pinitasd9eb2752018-04-03 13:44:29 +0100[diff] [blame]31using namespace arm_compute::graph::frontend;
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]32using namespace arm_compute::graph_utils;
33
34/** Example demonstrating how to implement InceptionV3's network using the Compute Library's graph API
35 *
36 * @param[in] argc Number of arguments
Georgios Pinitas12be7ab2018-07-03 12:06:23 +0100[diff] [blame]37 * @param[in] argv Arguments
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]38 */
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]39class InceptionV3Example : public Example
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]40{
41public:
Georgios Pinitas12be7ab2018-07-03 12:06:23 +0100[diff] [blame]42 InceptionV3Example()
43 : cmd_parser(), common_opts(cmd_parser), common_params(), graph(0, "InceptionV3")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]44 {
Georgios Pinitas12be7ab2018-07-03 12:06:23 +0100[diff] [blame]45 }
46 bool do_setup(int argc, char **argv) override
47 {
48 // Parse arguments
49 cmd_parser.parse(argc, argv);
50
51 // Consume common parameters
52 common_params = consume_common_graph_parameters(common_opts);
53
54 // Return when help menu is requested
55 if(common_params.help)
56 {
57 cmd_parser.print_help(argv[0]);
58 return false;
59 }
60
61 // Checks
Georgios Pinitas6ed43b52018-07-12 17:34:22 +0100[diff] [blame]62 ARM_COMPUTE_EXIT_ON_MSG(arm_compute::is_data_type_quantized_asymmetric(common_params.data_type), "Unsupported data type!");
63 ARM_COMPUTE_EXIT_ON_MSG(common_params.data_layout == DataLayout::NHWC, "Unsupported data layout!");
Georgios Pinitas12be7ab2018-07-03 12:06:23 +0100[diff] [blame]64
65 // Print parameter values
66 std::cout << common_params << std::endl;
67
68 // Get trainable parameters data path
69 std::string data_path = common_params.data_path;
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]70
Georgios Pinitas140fdc72018-02-16 11:42:38 +0000[diff] [blame]71 // Create a preprocessor object
72 std::unique_ptr<IPreprocessor> preprocessor = arm_compute::support::cpp14::make_unique<TFPreproccessor>();
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]73
Georgios Pinitas12be7ab2018-07-03 12:06:23 +0100[diff] [blame]74 graph << common_params.target
75 << common_params.fast_math_hint
76 << InputLayer(TensorDescriptor(TensorShape(299U, 299U, 3U, 1U), common_params.data_type),
77 get_input_accessor(common_params, std::move(preprocessor), false))
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]78 << ConvolutionLayer(3U, 3U, 32U,
79 get_weights_accessor(data_path, "/cnn_data/inceptionv3_model/Conv2d_1a_3x3_weights.npy"),
80 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(2, 2, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]81 .set_name("Conv2d_1a_3x3/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]82 << BatchNormalizationLayer(get_weights_accessor(data_path,
83 "/cnn_data/inceptionv3_model/Conv2d_1a_3x3_BatchNorm_moving_mean.npy"),
84 get_weights_accessor(data_path,
85 "/cnn_data/inceptionv3_model/Conv2d_1a_3x3_BatchNorm_moving_variance.npy"),
86 get_random_accessor(1.f, 1.f), get_weights_accessor(data_path,
87 "/cnn_data/inceptionv3_model/Conv2d_1a_3x3_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]88 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]89 .set_name("Conv2d_1a_3x3/BatchNorm/batchnorm")
90 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name("Conv2d_1a_3x3/Relu")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]91 << ConvolutionLayer(3U, 3U, 32U,
92 get_weights_accessor(data_path, "/cnn_data/inceptionv3_model/Conv2d_2a_3x3_weights.npy"),
93 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]94 .set_name("Conv2d_2a_3x3/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]95 << BatchNormalizationLayer(get_weights_accessor(data_path,
96 "/cnn_data/inceptionv3_model/Conv2d_2a_3x3_BatchNorm_moving_mean.npy"),
97 get_weights_accessor(data_path,
98 "/cnn_data/inceptionv3_model/Conv2d_2a_3x3_BatchNorm_moving_variance.npy"),
99 get_random_accessor(1.f, 1.f), get_weights_accessor(data_path,
100 "/cnn_data/inceptionv3_model/Conv2d_2a_3x3_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]101 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]102 .set_name("Conv2d_2a_3x3/BatchNorm/batchnorm")
103 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name("Conv2d_2a_3x3/Relu")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]104
105 << ConvolutionLayer(3U, 3U, 64U,
106 get_weights_accessor(data_path, "/cnn_data/inceptionv3_model/Conv2d_2b_3x3_weights.npy"),
107 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 1, 1))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]108 .set_name("Conv2d_2b_3x3/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]109 << BatchNormalizationLayer(get_weights_accessor(data_path,
110 "/cnn_data/inceptionv3_model/Conv2d_2b_3x3_BatchNorm_moving_mean.npy"),
111 get_weights_accessor(data_path,
112 "/cnn_data/inceptionv3_model/Conv2d_2b_3x3_BatchNorm_moving_variance.npy"),
113 get_random_accessor(1.f, 1.f), get_weights_accessor(data_path,
114 "/cnn_data/inceptionv3_model/Conv2d_2b_3x3_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]115 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]116 .set_name("Conv2d_2b_3x3/BatchNorm/batchnorm")
117 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name("Conv2d_2b_3x3/Relu")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]118
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]119 << PoolingLayer(PoolingLayerInfo(PoolingType::MAX, 3, PadStrideInfo(2, 2, 0, 0, DimensionRoundingType::CEIL))).set_name("MaxPool_3a_3x3/MaxPool")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]120
121 << ConvolutionLayer(1U, 1U, 80U,
122 get_weights_accessor(data_path, "/cnn_data/inceptionv3_model/Conv2d_3b_1x1_weights.npy"),
123 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]124 .set_name("Conv2d_3b_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]125 << BatchNormalizationLayer(get_weights_accessor(data_path,
126 "/cnn_data/inceptionv3_model/Conv2d_3b_1x1_BatchNorm_moving_mean.npy"),
127 get_weights_accessor(data_path,
128 "/cnn_data/inceptionv3_model/Conv2d_3b_1x1_BatchNorm_moving_variance.npy"),
129 get_random_accessor(1.f, 1.f), get_weights_accessor(data_path,
130 "/cnn_data/inceptionv3_model/Conv2d_3b_1x1_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]131 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]132 .set_name("Conv2d_3b_1x1/BatchNorm/batchnorm")
133 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name("Conv2d_3b_1x1/Relu")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]134
135 << ConvolutionLayer(3U, 3U, 192U,
136 get_weights_accessor(data_path, "/cnn_data/inceptionv3_model/Conv2d_4a_3x3_weights.npy"),
137 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]138 .set_name("Conv2d_4a_3x3/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]139 << BatchNormalizationLayer(get_weights_accessor(data_path,
140 "/cnn_data/inceptionv3_model/Conv2d_4a_3x3_BatchNorm_moving_mean.npy"),
141 get_weights_accessor(data_path,
142 "/cnn_data/inceptionv3_model/Conv2d_4a_3x3_BatchNorm_moving_variance.npy"),
143 get_random_accessor(1.f, 1.f), get_weights_accessor(data_path,
144 "/cnn_data/inceptionv3_model/Conv2d_4a_3x3_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]145 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]146 .set_name("Conv2d_4a_3x3/BatchNorm/batchnorm")
147 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name("Conv2d_4a_3x3/Relu")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]148
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]149 << PoolingLayer(PoolingLayerInfo(PoolingType::MAX, 3, PadStrideInfo(2, 2, 0, 0, DimensionRoundingType::CEIL))).set_name("MaxPool_5a_3x3/MaxPool");
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]150
Georgios Pinitas41c482d2018-04-17 13:23:26 +0100[diff] [blame]151 graph << get_inception_node_A(data_path, "Mixed_5b", 64U, std::make_tuple(48U, 64U), std::make_tuple(64U, 96U, 96U),
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]152 32U)
153 .set_name("Mixed_5b/concat");
Georgios Pinitas41c482d2018-04-17 13:23:26 +0100[diff] [blame]154 graph << get_inception_node_A(data_path, "Mixed_5c", 64U, std::make_tuple(48U, 64U), std::make_tuple(64U, 96U, 96U),
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]155 64U, true)
156 .set_name("Mixed_5c/concat");
Georgios Pinitas41c482d2018-04-17 13:23:26 +0100[diff] [blame]157 graph << get_inception_node_A(data_path, "Mixed_5d", 64U, std::make_tuple(48U, 64U), std::make_tuple(64U, 96U, 96U),
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]158 64U)
159 .set_name("Mixed_5d/concat");
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]160
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]161 graph << get_inception_node_B(data_path, "Mixed_6a", 384U, std::make_tuple(64U, 96U, 96U)).set_name("Mixed_6a/concat");
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]162
Georgios Pinitas41c482d2018-04-17 13:23:26 +0100[diff] [blame]163 graph << get_inception_node_C(data_path, "Mixed_6b", 192U, std::make_tuple(128U, 128U, 192U),
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]164 std::make_tuple(128U, 128U, 128U, 128U, 192U), 192U)
165 .set_name("Mixed_6b/concat");
Georgios Pinitas41c482d2018-04-17 13:23:26 +0100[diff] [blame]166 graph << get_inception_node_C(data_path, "Mixed_6c", 192U, std::make_tuple(160U, 160U, 192U),
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]167 std::make_tuple(160U, 160U, 160U, 160U, 192U), 192U)
168 .set_name("Mixed_6c/concat");
Georgios Pinitas41c482d2018-04-17 13:23:26 +0100[diff] [blame]169 graph << get_inception_node_C(data_path, "Mixed_6d", 192U, std::make_tuple(160U, 160U, 192U),
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]170 std::make_tuple(160U, 160U, 160U, 160U, 192U), 192U)
171 .set_name("Mixed_6d/concat");
Georgios Pinitas41c482d2018-04-17 13:23:26 +0100[diff] [blame]172 graph << get_inception_node_C(data_path, "Mixed_6e", 192U, std::make_tuple(192U, 192U, 192U),
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]173 std::make_tuple(192U, 192U, 192U, 192U, 192U), 192U)
174 .set_name("Mixed_6e/concat");
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]175
Georgios Pinitas41c482d2018-04-17 13:23:26 +0100[diff] [blame]176 graph << get_inception_node_D(data_path, "Mixed_7a", std::make_tuple(192U, 320U),
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]177 std::make_tuple(192U, 192U, 192U, 192U))
178 .set_name("Mixed_7a/concat");
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]179
Georgios Pinitas41c482d2018-04-17 13:23:26 +0100[diff] [blame]180 graph << get_inception_node_E(data_path, "Mixed_7b", 320U, std::make_tuple(384U, 384U, 384U),
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]181 std::make_tuple(448U, 384U, 384U, 384U), 192U)
182 .set_name("Mixed_7b/concat");
Georgios Pinitas41c482d2018-04-17 13:23:26 +0100[diff] [blame]183 graph << get_inception_node_E(data_path, "Mixed_7c", 320U, std::make_tuple(384U, 384U, 384U),
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]184 std::make_tuple(448U, 384U, 384U, 384U), 192U, true)
185 .set_name("Mixed_7c/concat");
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]186
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]187 graph << PoolingLayer(PoolingLayerInfo(PoolingType::AVG, 8, PadStrideInfo(1, 1, 0, 0, DimensionRoundingType::CEIL))).set_name("Logits/AvgPool_1a_8x8/AvgPool")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]188 << ConvolutionLayer(1U, 1U, 1001U, get_weights_accessor(data_path,
189 "/cnn_data/inceptionv3_model/Logits_Conv2d_1c_1x1_weights.npy"),
190 get_weights_accessor(data_path,
191 "/cnn_data/inceptionv3_model/Logits_Conv2d_1c_1x1_biases.npy"),
192 PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]193 .set_name("Logits/Conv2d_1c_1x1/convolution")
194 << ReshapeLayer(TensorShape(1001U)).set_name("Predictions/Reshape")
195 << SoftmaxLayer().set_name("Predictions/Softmax")
Georgios Pinitas12be7ab2018-07-03 12:06:23 +0100[diff] [blame]196 << OutputLayer(get_output_accessor(common_params, 5));
Gian Marcoc1b6e372018-02-21 18:03:26 +0000[diff] [blame]197
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]198 // Finalize graph
Georgios Pinitas9a8c6722018-03-21 17:52:35 +0000[diff] [blame]199 GraphConfig config;
Georgios Pinitas12be7ab2018-07-03 12:06:23 +0100[diff] [blame]200 config.num_threads = common_params.threads;
201 config.use_tuner = common_params.enable_tuner;
Anthony Barbier7b607dc2018-07-13 15:55:24 +0100[diff] [blame]202 config.tuner_file = common_params.tuner_file;
203
Georgios Pinitas12be7ab2018-07-03 12:06:23 +0100[diff] [blame]204 graph.finalize(common_params.target, config);
205
206 return true;
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]207 }
208
209 void do_run() override
210 {
211 graph.run();
212 }
213
214private:
Georgios Pinitas12be7ab2018-07-03 12:06:23 +0100[diff] [blame]215 CommandLineParser cmd_parser;
216 CommonGraphOptions common_opts;
217 CommonGraphParams common_params;
218 Stream graph;
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]219
220private:
221 BranchLayer get_inception_node_A(const std::string &data_path, std::string &&param_path,
222 unsigned int a_filt,
223 std::tuple<unsigned int, unsigned int> b_filters,
224 std::tuple<unsigned int, unsigned int, unsigned int> c_filters,
225 unsigned int d_filt,
226 bool is_name_different = false)
227 {
228 std::string total_path = "/cnn_data/inceptionv3_model/" + param_path + "_";
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]229
230 // This is due to a naming issue in the tf model
231 std::string conv_id0 = "_0a_";
232 std::string conv_id1 = "2d_0b_";
233 if(is_name_different)
234 {
235 conv_id0 = "_0b_";
236 conv_id1 = "_1_0c_";
237 }
238
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]239 SubStream i_a(graph);
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]240 i_a << ConvolutionLayer(
241 1U, 1U, a_filt,
242 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_weights.npy"),
243 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
244 PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]245 .set_name(param_path + "/Branch_0/Conv2d_0a_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]246 << BatchNormalizationLayer(
247 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
248 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
249 get_random_accessor(1.f, 1.f),
250 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]251 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]252 .set_name(param_path + "/Branch_0/Conv2d_0a_1x1/BatchNorm/batchnorm")
253 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_0/Conv2d_0a_1x1/Relu");
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]254
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]255 SubStream i_b(graph);
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]256 i_b << ConvolutionLayer(
257 1U, 1U, std::get<0>(b_filters),
258 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d" + conv_id0 + "1x1_weights.npy"),
259 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
260 PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]261 .set_name(param_path + "/Branch_1/Conv2d" + conv_id0 + "1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]262 << BatchNormalizationLayer(
263 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d" + conv_id0 + "1x1_BatchNorm_moving_mean.npy"),
264 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d" + conv_id0 + "1x1_BatchNorm_moving_variance.npy"),
265 get_random_accessor(1.f, 1.f),
266 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d" + conv_id0 + "1x1_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]267 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]268 .set_name(param_path + "/Branch_1/Conv2d" + conv_id0 + "1x1/BatchNorm/batchnorm")
269 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_1/Conv2d" + conv_id0 + "1x1/Relu")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]270 << ConvolutionLayer(
271 5U, 5U, std::get<1>(b_filters),
272 get_weights_accessor(data_path, total_path + "Branch_1_Conv" + conv_id1 + "5x5_weights.npy"),
273 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
274 PadStrideInfo(1, 1, 2, 2))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]275 .set_name(param_path + "/Branch_1/Conv2d" + conv_id1 + "5x5/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]276 << BatchNormalizationLayer(
277 get_weights_accessor(data_path, total_path + "Branch_1_Conv" + conv_id1 + "5x5_BatchNorm_moving_mean.npy"),
278 get_weights_accessor(data_path, total_path + "Branch_1_Conv" + conv_id1 + "5x5_BatchNorm_moving_variance.npy"),
279 get_random_accessor(1.f, 1.f),
280 get_weights_accessor(data_path, total_path + "Branch_1_Conv" + conv_id1 + "5x5_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]281 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]282 .set_name(param_path + "/Branch_1/Conv2d" + conv_id1 + "5x5/BatchNorm/batchnorm")
283 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_1/Conv2d" + conv_id1 + "5x5/Relu");
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]284
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]285 SubStream i_c(graph);
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]286 i_c << ConvolutionLayer(
287 1U, 1U, std::get<0>(c_filters),
288 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_weights.npy"),
289 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
290 PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]291 .set_name(param_path + "/Branch_2/Conv2d_0a_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]292 << BatchNormalizationLayer(
293 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
294 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
295 get_random_accessor(1.f, 1.f),
296 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]297 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]298 .set_name(param_path + "/Branch_2/Conv2d_0a_1x1/BatchNorm/batchnorm")
299 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_2/Conv2d_0a_1x1/Relu")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]300 << ConvolutionLayer(
301 3U, 3U, std::get<1>(c_filters),
302 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_3x3_weights.npy"),
303 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
304 PadStrideInfo(1, 1, 1, 1))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]305 .set_name(param_path + "/Branch_2/Conv2d_0b_3x3/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]306 << BatchNormalizationLayer(
307 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_3x3_BatchNorm_moving_mean.npy"),
308 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_3x3_BatchNorm_moving_variance.npy"),
309 get_random_accessor(1.f, 1.f),
310 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_3x3_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]311 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]312 .set_name(param_path + "/Branch_2/Conv2d_0b_3x3/BatchNorm/batchnorm")
313 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_2/Conv2d_0b_3x3/Relu")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]314 << ConvolutionLayer(
315 3U, 3U, std::get<2>(c_filters),
316 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_3x3_weights.npy"),
317 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
318 PadStrideInfo(1, 1, 1, 1))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]319 .set_name(param_path + "/Branch_2/Conv2d_0c_3x3/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]320 << BatchNormalizationLayer(
321 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_3x3_BatchNorm_moving_mean.npy"),
322 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_3x3_BatchNorm_moving_variance.npy"),
323 get_random_accessor(1.f, 1.f),
324 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_3x3_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]325 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]326 .set_name(param_path + "/Branch_2/Conv2d_0c_3x3/BatchNorm/batcnorm")
327 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_2/Conv2d_0c_3x3/Relu");
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]328
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]329 SubStream i_d(graph);
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]330 i_d << PoolingLayer(PoolingLayerInfo(PoolingType::AVG, 3, PadStrideInfo(1, 1, 1, 1, DimensionRoundingType::CEIL), true)).set_name(param_path + "/Branch_3/AvgPool_0a_3x3/AvgPool")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]331 << ConvolutionLayer(
332 1U, 1U, d_filt,
333 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_weights.npy"),
334 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
335 PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]336 .set_name(param_path + "/Branch_3/Conv2d_0b_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]337 << BatchNormalizationLayer(
338 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_BatchNorm_moving_mean.npy"),
339 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_BatchNorm_moving_variance.npy"),
340 get_random_accessor(1.f, 1.f),
341 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]342 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]343 .set_name(param_path + "/Branch_3/Conv2d_0b_1x1/BatchNorm/batchnorm")
344 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_3/Conv2d_0b_1x1/Relu");
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]345
346 return BranchLayer(BranchMergeMethod::DEPTH_CONCATENATE, std::move(i_a), std::move(i_b), std::move(i_c), std::move(i_d));
347 }
348
349 BranchLayer get_inception_node_B(const std::string &data_path, std::string &&param_path,
350 unsigned int a_filt,
351 std::tuple<unsigned int, unsigned int, unsigned int> b_filters)
352 {
353 std::string total_path = "/cnn_data/inceptionv3_model/" + param_path + "_";
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]354 SubStream i_a(graph);
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]355 i_a << ConvolutionLayer(
356 3U, 3U, a_filt,
357 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_1a_1x1_weights.npy"),
358 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
359 PadStrideInfo(2, 2, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]360 .set_name(param_path + "/Branch_0/Conv2d_1a_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]361 << BatchNormalizationLayer(
362 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_1a_1x1_BatchNorm_moving_mean.npy"),
363 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_1a_1x1_BatchNorm_moving_variance.npy"),
364 get_random_accessor(1.f, 1.f),
365 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_1a_1x1_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]366 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]367 .set_name(param_path + "/Branch_0/Conv2d_1a_1x1/BatchNorm/batchnorm")
368 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_0/Conv2d_1a_1x1/Relu");
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]369
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]370 SubStream i_b(graph);
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]371 i_b << ConvolutionLayer(
372 1U, 1U, std::get<0>(b_filters),
373 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_weights.npy"),
374 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
375 PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]376 .set_name(param_path + "/Branch_1/Conv2d_0a_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]377 << BatchNormalizationLayer(
378 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
379 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
380 get_random_accessor(1.f, 1.f),
381 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]382 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]383 .set_name(param_path + "/Branch_1/Conv2d_0a_1x1/BatchNorm/batchnorm")
384 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_1/Conv2d_0a_1x1/Relu")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]385 << ConvolutionLayer(
386 3U, 3U, std::get<1>(b_filters),
387 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_3x3_weights.npy"),
388 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
389 PadStrideInfo(1, 1, 1, 1))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]390 .set_name(param_path + "/Branch_1/Conv2d_0b_3x3/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]391 << BatchNormalizationLayer(
392 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_3x3_BatchNorm_moving_mean.npy"),
393 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_3x3_BatchNorm_moving_variance.npy"),
394 get_random_accessor(1.f, 1.f),
395 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_3x3_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]396 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]397 .set_name(param_path + "/Branch_1/Conv2d_0b_3x3/BatchNorm/batchnorm")
398 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_1/Conv2d_0b_3x3/Relu")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]399 << ConvolutionLayer(
400 3U, 3U, std::get<2>(b_filters),
401 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_1a_1x1_weights.npy"),
402 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
403 PadStrideInfo(2, 2, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]404 .set_name(param_path + "/Branch_1/Conv2d_1a_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]405 << BatchNormalizationLayer(
406 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_1a_1x1_BatchNorm_moving_mean.npy"),
407 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_1a_1x1_BatchNorm_moving_variance.npy"),
408 get_random_accessor(1.f, 1.f),
409 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_1a_1x1_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]410 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]411 .set_name(param_path + "/Branch_1/Conv2d_1a_1x1/BatchNorm/batchnorm")
412 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_1/Conv2d_1a_1x1/Relu");
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]413
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]414 SubStream i_c(graph);
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]415 i_c << PoolingLayer(PoolingLayerInfo(PoolingType::MAX, 3, PadStrideInfo(2, 2, 0, 0, DimensionRoundingType::CEIL))).set_name(param_path + "/Branch_2/MaxPool_1a_3x3/MaxPool");
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]416
417 return BranchLayer(BranchMergeMethod::DEPTH_CONCATENATE, std::move(i_a), std::move(i_b), std::move(i_c));
418 }
419
420 BranchLayer get_inception_node_C(const std::string &data_path, std::string &&param_path,
421 unsigned int a_filt,
422 std::tuple<unsigned int, unsigned int, unsigned int> b_filters,
423 std::tuple<unsigned int, unsigned int, unsigned int, unsigned int, unsigned int> c_filters,
424 unsigned int d_filt)
425 {
426 std::string total_path = "/cnn_data/inceptionv3_model/" + param_path + "_";
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]427 SubStream i_a(graph);
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]428 i_a << ConvolutionLayer(
429 1U, 1U, a_filt,
430 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_weights.npy"),
431 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
432 PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]433 .set_name(param_path + "/Branch_0/Conv2d_0a_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]434 << BatchNormalizationLayer(
435 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
436 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
437 get_random_accessor(1.f, 1.f),
438 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]439 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]440 .set_name(param_path + "/Branch_0/Conv2d_0a_1x1/BatchNorm/batchnorm")
441 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_0/Conv2d_0a_1x1/Relu");
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]442
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]443 SubStream i_b(graph);
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]444 i_b << ConvolutionLayer(
445 1U, 1U, std::get<0>(b_filters),
446 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_weights.npy"),
447 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
448 PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]449 .set_name(param_path + "/Branch_1/Conv2d_0a_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]450 << BatchNormalizationLayer(
451 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
452 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
453 get_random_accessor(1.f, 1.f),
454 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]455 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]456 .set_name(param_path + "/Branch_1/Conv2d_0a_1x1/BatchNorm/batchnorm")
457 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_1/Conv2d_0a_1x1/Relu")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]458 << ConvolutionLayer(
459 7U, 1U, std::get<1>(b_filters),
460 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x7_weights.npy"),
461 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
462 PadStrideInfo(1, 1, 3, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]463 .set_name(param_path + "/Branch_1/Conv2d_0b_1x7/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]464 << BatchNormalizationLayer(
465 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x7_BatchNorm_moving_mean.npy"),
466 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x7_BatchNorm_moving_variance.npy"),
467 get_random_accessor(1.f, 1.f),
468 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x7_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]469 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]470 .set_name(param_path + "/Branch_1/Conv2d_0b_1x7/BatchNorm/batchnorm")
471 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_1/Conv2d_0b_1x7/Relu")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]472 << ConvolutionLayer(
473 1U, 7U, std::get<2>(b_filters),
474 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0c_7x1_weights.npy"),
475 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
476 PadStrideInfo(1, 1, 0, 3))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]477 .set_name(param_path + "/Branch_1/Conv2d_0c_7x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]478 << BatchNormalizationLayer(
479 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0c_7x1_BatchNorm_moving_mean.npy"),
480 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0c_7x1_BatchNorm_moving_variance.npy"),
481 get_random_accessor(1.f, 1.f),
482 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0c_7x1_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]483 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]484 .set_name(param_path + "/Branch_1/Conv2d_0c_7x1/BatchNorm/batchnorm")
485 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_0/Conv2d_0c_7x1/Relu");
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]486
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]487 SubStream i_c(graph);
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]488 i_c << ConvolutionLayer(
489 1U, 1U, std::get<0>(c_filters),
490 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_weights.npy"),
491 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
492 PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]493 .set_name(param_path + "/Branch_2/Conv2d_0a_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]494 << BatchNormalizationLayer(
495 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
496 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
497 get_random_accessor(1.f, 1.f),
498 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]499 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]500 .set_name(param_path + "/Branch_2/Conv2d_0a_1x1/BatchNorm/batchnorm")
501 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_2/Conv2d_0a_1x1/Relu")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]502 << ConvolutionLayer(
503 1U, 7U, std::get<1>(c_filters),
504 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_7x1_weights.npy"),
505 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
506 PadStrideInfo(1, 1, 0, 3))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]507 .set_name(param_path + "/Branch_2/Conv2d_0b_7x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]508 << BatchNormalizationLayer(
509 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_7x1_BatchNorm_moving_mean.npy"),
510 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_7x1_BatchNorm_moving_variance.npy"),
511 get_random_accessor(1.f, 1.f),
512 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_7x1_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]513 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]514 .set_name(param_path + "/Branch_2/Conv2d_0b_7x1/BatchNorm/batchnorm")
515 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_2/Conv2d_0b_7x1/Relu")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]516 << ConvolutionLayer(
517 7U, 1U, std::get<2>(c_filters),
518 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_1x7_weights.npy"),
519 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
520 PadStrideInfo(1, 1, 3, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]521 .set_name(param_path + "/Branch_2/Conv2d_0c_1x7/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]522 << BatchNormalizationLayer(
523 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_1x7_BatchNorm_moving_mean.npy"),
524 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_1x7_BatchNorm_moving_variance.npy"),
525 get_random_accessor(1.f, 1.f),
526 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_1x7_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]527 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]528 .set_name(param_path + "/Branch_2/Conv2d_0c_1x7/BatchNorm/batchnorm")
529 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_2/Conv2d_0c_1x7/Relu")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]530 << ConvolutionLayer(
531 1U, 7U, std::get<3>(c_filters),
532 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0d_7x1_weights.npy"),
533 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
534 PadStrideInfo(1, 1, 0, 3))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]535 .set_name(param_path + "/Branch_2/Conv2d_0d_7x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]536 << BatchNormalizationLayer(
537 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0d_7x1_BatchNorm_moving_mean.npy"),
538 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0d_7x1_BatchNorm_moving_variance.npy"),
539 get_random_accessor(1.f, 1.f),
540 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0d_7x1_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]541 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]542 .set_name(param_path + "/Branch_2/Conv2d_0d_7x1/BatchNorm/batchnorm")
543 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_2/Conv2d_0d_7x1/Relu")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]544 << ConvolutionLayer(
545 7U, 1U, std::get<4>(c_filters),
546 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0e_1x7_weights.npy"),
547 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
548 PadStrideInfo(1, 1, 3, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]549 .set_name(param_path + "/Branch_2/Conv2d_0e_1x7/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]550 << BatchNormalizationLayer(
551 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0e_1x7_BatchNorm_moving_mean.npy"),
552 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0e_1x7_BatchNorm_moving_variance.npy"),
553 get_random_accessor(1.f, 1.f),
554 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0e_1x7_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]555 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]556 .set_name(param_path + "/Branch_2/Conv2d_0e_1x7/BatchNorm/batchnorm")
557 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_2/Conv2d_0e_1x7/Relu");
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]558
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]559 SubStream i_d(graph);
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]560 i_d << PoolingLayer(PoolingLayerInfo(PoolingType::AVG, 3, PadStrideInfo(1, 1, 1, 1, DimensionRoundingType::CEIL), true)).set_name(param_path + "/Branch_3/AvgPool_0a_3x3/AvgPool")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]561 << ConvolutionLayer(
562 1U, 1U, d_filt,
563 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_weights.npy"),
564 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
565 PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]566 .set_name(param_path + "/Branch_3/Conv2d_0b_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]567 << BatchNormalizationLayer(
568 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_BatchNorm_moving_mean.npy"),
569 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_BatchNorm_moving_variance.npy"),
570 get_random_accessor(1.f, 1.f),
571 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]572 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]573 .set_name(param_path + "/Branch_3/Conv2d_0b_1x1/BatchNorm/batchnorm")
574 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_3/Conv2d_0b_1x1/Relu");
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]575
576 return BranchLayer(BranchMergeMethod::DEPTH_CONCATENATE, std::move(i_a), std::move(i_b), std::move(i_c), std::move(i_d));
577 }
578
579 BranchLayer get_inception_node_D(const std::string &data_path, std::string &&param_path,
580 std::tuple<unsigned int, unsigned int> a_filters,
581 std::tuple<unsigned int, unsigned int, unsigned int, unsigned int> b_filters)
582 {
583 std::string total_path = "/cnn_data/inceptionv3_model/" + param_path + "_";
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]584 SubStream i_a(graph);
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]585 i_a << ConvolutionLayer(
586 1U, 1U, std::get<0>(a_filters),
587 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_weights.npy"),
588 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
589 PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]590 .set_name(param_path + "/Branch_0/Conv2d_0a_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]591 << BatchNormalizationLayer(
592 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
593 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
594 get_random_accessor(1.f, 1.f),
595 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]596 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]597 .set_name(param_path + "/Branch_0/Conv2d_0a_1x1/BatchNorm/batchnorm")
598 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_0/Conv2d_0a_1x1/Relu")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]599 << ConvolutionLayer(
600 3U, 3U, std::get<1>(a_filters),
601 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_1a_3x3_weights.npy"),
602 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
603 PadStrideInfo(2, 2, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]604 .set_name(param_path + "/Branch_0/Conv2d_1a_3x3/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]605 << BatchNormalizationLayer(
606 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_1a_3x3_BatchNorm_moving_mean.npy"),
607 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_1a_3x3_BatchNorm_moving_variance.npy"),
608 get_random_accessor(1.f, 1.f),
609 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_1a_3x3_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]610 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]611 .set_name(param_path + "/Branch_0/Conv2d_1a_3x3/BatchNorm/batchnorm")
612 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_0/Conv2d_1a_3x3/Relu");
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]613
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]614 SubStream i_b(graph);
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]615 i_b << ConvolutionLayer(
616 1U, 1U, std::get<0>(b_filters),
617 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_weights.npy"),
618 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
619 PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]620 .set_name(param_path + "/Branch_1/Conv2d_0a_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]621 << BatchNormalizationLayer(
622 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
623 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
624 get_random_accessor(1.f, 1.f),
625 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]626 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]627 .set_name(param_path + "/Branch_1/Conv2d_0a_1x1/BatchNorm/batchnorm")
628 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_1/Conv2d_0a_1x1/Relu")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]629 << ConvolutionLayer(
630 7U, 1U, std::get<1>(b_filters),
631 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x7_weights.npy"),
632 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
633 PadStrideInfo(1, 1, 3, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]634 .set_name(param_path + "/Branch_1/Conv2d_0b_1x7/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]635 << BatchNormalizationLayer(
636 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x7_BatchNorm_moving_mean.npy"),
637 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x7_BatchNorm_moving_variance.npy"),
638 get_random_accessor(1.f, 1.f),
639 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x7_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]640 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]641 .set_name(param_path + "/Branch_1/Conv2d_0b_1x7/BatchNorm/batchnorm")
642 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_1/Conv2d_0b_1x7/Relu")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]643 << ConvolutionLayer(
644 1U, 7U, std::get<2>(b_filters),
645 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0c_7x1_weights.npy"),
646 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
647 PadStrideInfo(1, 1, 0, 3))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]648 .set_name(param_path + "/Branch_1/Conv2d_0c_7x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]649 << BatchNormalizationLayer(
650 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0c_7x1_BatchNorm_moving_mean.npy"),
651 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0c_7x1_BatchNorm_moving_variance.npy"),
652 get_random_accessor(1.f, 1.f),
653 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0c_7x1_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]654 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]655 .set_name(param_path + "/Branch_1/Conv2d_0c_7x1/BatchNorm/batchnorm")
656 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_1/Conv2d_0c_7x1/Relu")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]657 << ConvolutionLayer(
658 3U, 3U, std::get<3>(b_filters),
659 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_1a_3x3_weights.npy"),
660 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
661 PadStrideInfo(2, 2, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]662 .set_name(param_path + "/Branch_1/Conv2d_1a_3x3/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]663 << BatchNormalizationLayer(
664 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_1a_3x3_BatchNorm_moving_mean.npy"),
665 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_1a_3x3_BatchNorm_moving_variance.npy"),
666 get_random_accessor(1.f, 1.f),
667 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_1a_3x3_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]668 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]669 .set_name(param_path + "/Branch_1/Conv2d_1a_3x3/BatchNorm/batchnorm")
670 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_1/Conv2d_1a_3x3/Relu");
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]671
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]672 SubStream i_c(graph);
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]673 i_c << PoolingLayer(PoolingLayerInfo(PoolingType::MAX, 3, PadStrideInfo(2, 2, 0, 0, DimensionRoundingType::CEIL))).set_name(param_path + "/Branch_2/MaxPool_1a_3x3/MaxPool");
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]674
675 return BranchLayer(BranchMergeMethod::DEPTH_CONCATENATE, std::move(i_a), std::move(i_b), std::move(i_c));
676 }
677
678 BranchLayer get_inception_node_E(const std::string &data_path, std::string &&param_path,
679 unsigned int a_filt,
680 std::tuple<unsigned int, unsigned int, unsigned int> b_filters,
681 std::tuple<unsigned int, unsigned int, unsigned int, unsigned int> c_filters,
682 unsigned int d_filt,
683 bool is_name_different = false)
684 {
685 // This is due to a naming issue in the tf model
686 std::string conv_id = "_0b_";
687 if(is_name_different)
688 {
689 conv_id = "_0c_";
690 }
691
692 std::string total_path = "/cnn_data/inceptionv3_model/" + param_path + "_";
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]693 SubStream i_a(graph);
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]694 i_a << ConvolutionLayer(
695 1U, 1U, a_filt,
696 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_weights.npy"),
697 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
698 PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]699 .set_name(param_path + "/Branch_0/Conv2d_0a_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]700 << BatchNormalizationLayer(
701 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
702 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
703 get_random_accessor(1.f, 1.f),
704 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]705 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]706 .set_name(param_path + "/Branch_0/Conv2d_0a_1x1/BatchNorm/batchnorm")
707 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_0/Conv2d_0a_1x1/Relu");
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]708
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]709 SubStream i_b(graph);
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]710 i_b << ConvolutionLayer(
711 1U, 1U, std::get<0>(b_filters),
712 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_weights.npy"),
713 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
714 PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]715 .set_name(param_path + "/Branch_1/Conv2d_0a_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]716 << BatchNormalizationLayer(
717 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
718 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
719 get_random_accessor(1.f, 1.f),
720 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]721 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]722 .set_name(param_path + "/Branch_1/Conv2d_0a_1x1/BatchNorm/batchnorm")
723 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_1/Conv2d_0a_1x1/Relu");
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]724
Georgios Pinitas772e17f2018-07-13 12:25:33 +0100[diff] [blame]725 SubStream i_b1(i_b);
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]726 i_b1 << ConvolutionLayer(
727 3U, 1U, std::get<1>(b_filters),
728 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x3_weights.npy"),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]729 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
730 PadStrideInfo(1, 1, 1, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]731 .set_name(param_path + "/Branch_1/Conv2d_0b_1x3/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]732 << BatchNormalizationLayer(
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]733 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x3_BatchNorm_moving_mean.npy"),
734 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x3_BatchNorm_moving_variance.npy"),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]735 get_random_accessor(1.f, 1.f),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]736 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x3_BatchNorm_beta.npy"),
737 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]738 .set_name(param_path + "/Branch_1/Conv2d_0b_1x3/BatchNorm/batchnorm")
739 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_1/Conv2d_0b_1x3/Relu");
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]740
Georgios Pinitas772e17f2018-07-13 12:25:33 +0100[diff] [blame]741 SubStream i_b2(i_b);
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]742 i_b2 << ConvolutionLayer(
743 1U, 3U, std::get<2>(b_filters),
744 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d" + conv_id + "3x1_weights.npy"),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]745 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
746 PadStrideInfo(1, 1, 0, 1))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]747 .set_name(param_path + "/Branch_1/Conv2d" + conv_id + "3x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]748 << BatchNormalizationLayer(
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]749 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d" + conv_id + "3x1_BatchNorm_moving_mean.npy"),
750 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d" + conv_id + "3x1_BatchNorm_moving_variance.npy"),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]751 get_random_accessor(1.f, 1.f),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]752 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d" + conv_id + "3x1_BatchNorm_beta.npy"),
753 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]754 .set_name(param_path + "/Branch_1/Conv2d" + conv_id + "3x1/BatchNorm/batchnorm")
755 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_1/Conv2d" + conv_id + "3x1/Relu");
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]756
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]757 // Merge b1 and b2
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]758 i_b << BranchLayer(BranchMergeMethod::DEPTH_CONCATENATE, std::move(i_b1), std::move(i_b2)).set_name(param_path + "/Branch_1/concat");
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]759
760 SubStream i_c(graph);
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]761 i_c << ConvolutionLayer(
762 1U, 1U, std::get<0>(c_filters),
763 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_weights.npy"),
764 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
765 PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]766 .set_name(param_path + "/Branch_2/Conv2d_0a_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]767 << BatchNormalizationLayer(
768 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
769 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
770 get_random_accessor(1.f, 1.f),
771 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]772 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]773 .set_name(param_path + "/Branch_2/Conv2d_0a_1x1/BatchNorm/batchnorm")
774 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_2/Conv2d_0a_1x1/Relu")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]775 << ConvolutionLayer(
776 3U, 3U, std::get<1>(c_filters),
777 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_3x3_weights.npy"),
778 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
779 PadStrideInfo(1, 1, 1, 1))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]780 .set_name(param_path + "/Branch_2/Conv2d_0b_3x3/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]781 << BatchNormalizationLayer(
782 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_3x3_BatchNorm_moving_mean.npy"),
783 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_3x3_BatchNorm_moving_variance.npy"),
784 get_random_accessor(1.f, 1.f),
785 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_3x3_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]786 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]787 .set_name(param_path + "/Branch_2/Conv2d_0b_3x3/BatchNorm/batchnorm")
788 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_2/Conv2d_0b_3x3/Relu");
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]789
Georgios Pinitas772e17f2018-07-13 12:25:33 +0100[diff] [blame]790 SubStream i_c1(i_c);
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]791 i_c1 << ConvolutionLayer(
792 3U, 1U, std::get<2>(c_filters),
793 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_1x3_weights.npy"),
794 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
795 PadStrideInfo(1, 1, 1, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]796 .set_name(param_path + "/Branch_2/Conv2d_0c_1x3/convolution")
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]797 << BatchNormalizationLayer(
798 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_1x3_BatchNorm_moving_mean.npy"),
799 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_1x3_BatchNorm_moving_variance.npy"),
800 get_random_accessor(1.f, 1.f),
801 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_1x3_BatchNorm_beta.npy"),
802 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]803 .set_name(param_path + "/Branch_2/Conv2d_0c_1x3/BatchNorm/batchnorm")
804 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_2/Conv2d_0c_1x3/Relu");
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]805
Georgios Pinitas772e17f2018-07-13 12:25:33 +0100[diff] [blame]806 SubStream i_c2(i_c);
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]807 i_c2 << ConvolutionLayer(
808 1U, 3U, std::get<3>(c_filters),
809 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0d_3x1_weights.npy"),
810 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
811 PadStrideInfo(1, 1, 0, 1))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]812 .set_name(param_path + "/Branch_2/Conv2d_0d_3x1/convolution")
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]813 << BatchNormalizationLayer(
814 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0d_3x1_BatchNorm_moving_mean.npy"),
815 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0d_3x1_BatchNorm_moving_variance.npy"),
816 get_random_accessor(1.f, 1.f),
817 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0d_3x1_BatchNorm_beta.npy"),
818 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]819 .set_name(param_path + "/Branch_2/Conv2d_0d_3x1/BatchNorm/batchnorm")
820 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_2/Conv2d_0d_3x1/Relu");
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]821
822 // Merge i_c1 and i_c2
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]823 i_c << BranchLayer(BranchMergeMethod::DEPTH_CONCATENATE, std::move(i_c1), std::move(i_c2)).set_name(param_path + "/Branch_2/concat");
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]824
825 SubStream i_d(graph);
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]826 i_d << PoolingLayer(PoolingLayerInfo(PoolingType::AVG, 3, PadStrideInfo(1, 1, 1, 1, DimensionRoundingType::CEIL), true)).set_name(param_path + "/Branch_3/AvgPool_0a_3x3/AvgPool")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]827 << ConvolutionLayer(
828 1U, 1U, d_filt,
829 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_weights.npy"),
830 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
831 PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]832 .set_name(param_path + "/Branch_3/Conv2d_0b_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]833 << BatchNormalizationLayer(
834 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_BatchNorm_moving_mean.npy"),
835 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_BatchNorm_moving_variance.npy"),
836 get_random_accessor(1.f, 1.f),
837 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]838 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]839 .set_name(param_path + "/Branch_3/Conv2d_0b_1x1/BatchNorm/batchnorm")
840 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name(param_path + "/Branch_3/Conv2d_0b_1x1/Relu");
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]841
842 return BranchLayer(BranchMergeMethod::DEPTH_CONCATENATE, std::move(i_a), std::move(i_b), std::move(i_c), std::move(i_d));
843 }
844};
845
846/** Main program for Inception V3
847 *
Georgios Pinitas9f28b392018-07-18 20:01:53 +0100[diff] [blame]848 * @note To list all the possible arguments execute the binary appended with the --help option
849 *
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]850 * @param[in] argc Number of arguments
Georgios Pinitas12be7ab2018-07-03 12:06:23 +0100[diff] [blame]851 * @param[in] argv Arguments
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]852 */
853int main(int argc, char **argv)
854{
855 return arm_compute::utils::run_example<InceptionV3Example>(argc, argv);
856}