Gitiles
Code Review Sign In
review.mlplatform.org / ml / ComputeLibrary / 17b0f8ba60ec9db4b96471f9406843bee6a43a4f / . / examples / graph_inception_v3.cpp
blob: 86b223e9e54f45ea26377428f056e7bf309ad95d [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
Georgios Pinitas108ab0b2018-09-14 18:35:11 +0100[diff] [blame]34/** Example demonstrating how to implement InceptionV3's network using the Compute Library's graph API */
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]35class InceptionV3Example : public Example
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]36{
37public:
Georgios Pinitas12be7ab2018-07-03 12:06:23 +0100[diff] [blame]38 InceptionV3Example()
39 : cmd_parser(), common_opts(cmd_parser), common_params(), graph(0, "InceptionV3")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]40 {
Georgios Pinitas12be7ab2018-07-03 12:06:23 +0100[diff] [blame]41 }
42 bool do_setup(int argc, char **argv) override
43 {
44 // Parse arguments
45 cmd_parser.parse(argc, argv);
46
47 // Consume common parameters
48 common_params = consume_common_graph_parameters(common_opts);
49
50 // Return when help menu is requested
51 if(common_params.help)
52 {
53 cmd_parser.print_help(argv[0]);
54 return false;
55 }
56
57 // Checks
Anthony Barbiercdd68c02018-08-23 15:03:41 +0100[diff] [blame]58 ARM_COMPUTE_EXIT_ON_MSG(arm_compute::is_data_type_quantized_asymmetric(common_params.data_type), "QASYMM8 not supported for this graph");
Georgios Pinitas12be7ab2018-07-03 12:06:23 +0100[diff] [blame]59
60 // Print parameter values
61 std::cout << common_params << std::endl;
62
63 // Get trainable parameters data path
64 std::string data_path = common_params.data_path;
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]65
Georgios Pinitas140fdc72018-02-16 11:42:38 +0000[diff] [blame]66 // Create a preprocessor object
67 std::unique_ptr<IPreprocessor> preprocessor = arm_compute::support::cpp14::make_unique<TFPreproccessor>();
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]68
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]69 // Create input descriptor
70 const TensorShape tensor_shape = permute_shape(TensorShape(299U, 299U, 3U, 1U), DataLayout::NCHW, common_params.data_layout);
71 TensorDescriptor input_descriptor = TensorDescriptor(tensor_shape, common_params.data_type).set_layout(common_params.data_layout);
72
73 // Set weights trained layout
74 const DataLayout weights_layout = DataLayout::NCHW;
75
Georgios Pinitas12be7ab2018-07-03 12:06:23 +0100[diff] [blame]76 graph << common_params.target
77 << common_params.fast_math_hint
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]78 << InputLayer(input_descriptor, get_input_accessor(common_params, std::move(preprocessor), false))
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]79 << ConvolutionLayer(3U, 3U, 32U,
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]80 get_weights_accessor(data_path, "/cnn_data/inceptionv3_model/Conv2d_1a_3x3_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]81 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(2, 2, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]82 .set_name("Conv2d_1a_3x3/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]83 << BatchNormalizationLayer(get_weights_accessor(data_path,
84 "/cnn_data/inceptionv3_model/Conv2d_1a_3x3_BatchNorm_moving_mean.npy"),
85 get_weights_accessor(data_path,
86 "/cnn_data/inceptionv3_model/Conv2d_1a_3x3_BatchNorm_moving_variance.npy"),
87 get_random_accessor(1.f, 1.f), get_weights_accessor(data_path,
88 "/cnn_data/inceptionv3_model/Conv2d_1a_3x3_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]89 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]90 .set_name("Conv2d_1a_3x3/BatchNorm/batchnorm")
91 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name("Conv2d_1a_3x3/Relu")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]92 << ConvolutionLayer(3U, 3U, 32U,
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]93 get_weights_accessor(data_path, "/cnn_data/inceptionv3_model/Conv2d_2a_3x3_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]94 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]95 .set_name("Conv2d_2a_3x3/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]96 << BatchNormalizationLayer(get_weights_accessor(data_path,
97 "/cnn_data/inceptionv3_model/Conv2d_2a_3x3_BatchNorm_moving_mean.npy"),
98 get_weights_accessor(data_path,
99 "/cnn_data/inceptionv3_model/Conv2d_2a_3x3_BatchNorm_moving_variance.npy"),
100 get_random_accessor(1.f, 1.f), get_weights_accessor(data_path,
101 "/cnn_data/inceptionv3_model/Conv2d_2a_3x3_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]102 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]103 .set_name("Conv2d_2a_3x3/BatchNorm/batchnorm")
104 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name("Conv2d_2a_3x3/Relu")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]105
106 << ConvolutionLayer(3U, 3U, 64U,
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]107 get_weights_accessor(data_path, "/cnn_data/inceptionv3_model/Conv2d_2b_3x3_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]108 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 1, 1))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]109 .set_name("Conv2d_2b_3x3/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]110 << BatchNormalizationLayer(get_weights_accessor(data_path,
111 "/cnn_data/inceptionv3_model/Conv2d_2b_3x3_BatchNorm_moving_mean.npy"),
112 get_weights_accessor(data_path,
113 "/cnn_data/inceptionv3_model/Conv2d_2b_3x3_BatchNorm_moving_variance.npy"),
114 get_random_accessor(1.f, 1.f), get_weights_accessor(data_path,
115 "/cnn_data/inceptionv3_model/Conv2d_2b_3x3_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]116 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]117 .set_name("Conv2d_2b_3x3/BatchNorm/batchnorm")
118 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name("Conv2d_2b_3x3/Relu")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]119
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]120 << 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]121
122 << ConvolutionLayer(1U, 1U, 80U,
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]123 get_weights_accessor(data_path, "/cnn_data/inceptionv3_model/Conv2d_3b_1x1_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]124 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]125 .set_name("Conv2d_3b_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]126 << BatchNormalizationLayer(get_weights_accessor(data_path,
127 "/cnn_data/inceptionv3_model/Conv2d_3b_1x1_BatchNorm_moving_mean.npy"),
128 get_weights_accessor(data_path,
129 "/cnn_data/inceptionv3_model/Conv2d_3b_1x1_BatchNorm_moving_variance.npy"),
130 get_random_accessor(1.f, 1.f), get_weights_accessor(data_path,
131 "/cnn_data/inceptionv3_model/Conv2d_3b_1x1_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]132 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]133 .set_name("Conv2d_3b_1x1/BatchNorm/batchnorm")
134 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name("Conv2d_3b_1x1/Relu")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]135
136 << ConvolutionLayer(3U, 3U, 192U,
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]137 get_weights_accessor(data_path, "/cnn_data/inceptionv3_model/Conv2d_4a_3x3_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]138 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr), PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]139 .set_name("Conv2d_4a_3x3/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]140 << BatchNormalizationLayer(get_weights_accessor(data_path,
141 "/cnn_data/inceptionv3_model/Conv2d_4a_3x3_BatchNorm_moving_mean.npy"),
142 get_weights_accessor(data_path,
143 "/cnn_data/inceptionv3_model/Conv2d_4a_3x3_BatchNorm_moving_variance.npy"),
144 get_random_accessor(1.f, 1.f), get_weights_accessor(data_path,
145 "/cnn_data/inceptionv3_model/Conv2d_4a_3x3_BatchNorm_beta.npy"),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]146 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]147 .set_name("Conv2d_4a_3x3/BatchNorm/batchnorm")
148 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)).set_name("Conv2d_4a_3x3/Relu")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]149
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]150 << 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]151
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]152 graph << get_inception_node_A(data_path, "Mixed_5b", weights_layout, 64U, std::make_tuple(48U, 64U), std::make_tuple(64U, 96U, 96U),
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]153 32U)
154 .set_name("Mixed_5b/concat");
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]155 graph << get_inception_node_A(data_path, "Mixed_5c", weights_layout, 64U, std::make_tuple(48U, 64U), std::make_tuple(64U, 96U, 96U),
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]156 64U, true)
157 .set_name("Mixed_5c/concat");
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]158 graph << get_inception_node_A(data_path, "Mixed_5d", weights_layout, 64U, std::make_tuple(48U, 64U), std::make_tuple(64U, 96U, 96U),
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]159 64U)
160 .set_name("Mixed_5d/concat");
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]161
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]162 graph << get_inception_node_B(data_path, "Mixed_6a", weights_layout, 384U, std::make_tuple(64U, 96U, 96U)).set_name("Mixed_6a/concat");
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]163
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]164 graph << get_inception_node_C(data_path, "Mixed_6b", weights_layout, 192U, std::make_tuple(128U, 128U, 192U),
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]165 std::make_tuple(128U, 128U, 128U, 128U, 192U), 192U)
166 .set_name("Mixed_6b/concat");
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]167 graph << get_inception_node_C(data_path, "Mixed_6c", weights_layout, 192U, std::make_tuple(160U, 160U, 192U),
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]168 std::make_tuple(160U, 160U, 160U, 160U, 192U), 192U)
169 .set_name("Mixed_6c/concat");
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]170 graph << get_inception_node_C(data_path, "Mixed_6d", weights_layout, 192U, std::make_tuple(160U, 160U, 192U),
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]171 std::make_tuple(160U, 160U, 160U, 160U, 192U), 192U)
172 .set_name("Mixed_6d/concat");
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]173 graph << get_inception_node_C(data_path, "Mixed_6e", weights_layout, 192U, std::make_tuple(192U, 192U, 192U),
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]174 std::make_tuple(192U, 192U, 192U, 192U, 192U), 192U)
175 .set_name("Mixed_6e/concat");
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]176
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]177 graph << get_inception_node_D(data_path, "Mixed_7a", weights_layout, std::make_tuple(192U, 320U),
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]178 std::make_tuple(192U, 192U, 192U, 192U))
179 .set_name("Mixed_7a/concat");
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]180
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]181 graph << get_inception_node_E(data_path, "Mixed_7b", weights_layout, 320U, std::make_tuple(384U, 384U, 384U),
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]182 std::make_tuple(448U, 384U, 384U, 384U), 192U)
183 .set_name("Mixed_7b/concat");
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]184 graph << get_inception_node_E(data_path, "Mixed_7c", weights_layout, 320U, std::make_tuple(384U, 384U, 384U),
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]185 std::make_tuple(448U, 384U, 384U, 384U), 192U, true)
186 .set_name("Mixed_7c/concat");
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]187
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]188 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]189 << ConvolutionLayer(1U, 1U, 1001U, get_weights_accessor(data_path,
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]190 "/cnn_data/inceptionv3_model/Logits_Conv2d_1c_1x1_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]191 get_weights_accessor(data_path,
192 "/cnn_data/inceptionv3_model/Logits_Conv2d_1c_1x1_biases.npy"),
193 PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]194 .set_name("Logits/Conv2d_1c_1x1/convolution")
195 << ReshapeLayer(TensorShape(1001U)).set_name("Predictions/Reshape")
196 << SoftmaxLayer().set_name("Predictions/Softmax")
Georgios Pinitas12be7ab2018-07-03 12:06:23 +0100[diff] [blame]197 << OutputLayer(get_output_accessor(common_params, 5));
Gian Marcoc1b6e372018-02-21 18:03:26 +0000[diff] [blame]198
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]199 // Finalize graph
Georgios Pinitas9a8c6722018-03-21 17:52:35 +0000[diff] [blame]200 GraphConfig config;
Georgios Pinitas12be7ab2018-07-03 12:06:23 +0100[diff] [blame]201 config.num_threads = common_params.threads;
202 config.use_tuner = common_params.enable_tuner;
Anthony Barbier7b607dc2018-07-13 15:55:24 +0100[diff] [blame]203 config.tuner_file = common_params.tuner_file;
204
Georgios Pinitas12be7ab2018-07-03 12:06:23 +0100[diff] [blame]205 graph.finalize(common_params.target, config);
206
207 return true;
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]208 }
209
210 void do_run() override
211 {
212 graph.run();
213 }
214
215private:
Georgios Pinitas12be7ab2018-07-03 12:06:23 +0100[diff] [blame]216 CommandLineParser cmd_parser;
217 CommonGraphOptions common_opts;
218 CommonGraphParams common_params;
219 Stream graph;
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]220
221private:
Georgios Pinitas427bbbf2018-08-28 13:32:02 +0100[diff] [blame]222 ConcatLayer get_inception_node_A(const std::string &data_path, std::string &&param_path, DataLayout weights_layout,
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]223 unsigned int a_filt,
224 std::tuple<unsigned int, unsigned int> b_filters,
225 std::tuple<unsigned int, unsigned int, unsigned int> c_filters,
226 unsigned int d_filt,
227 bool is_name_different = false)
228 {
229 std::string total_path = "/cnn_data/inceptionv3_model/" + param_path + "_";
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]230
231 // This is due to a naming issue in the tf model
232 std::string conv_id0 = "_0a_";
233 std::string conv_id1 = "2d_0b_";
234 if(is_name_different)
235 {
236 conv_id0 = "_0b_";
237 conv_id1 = "_1_0c_";
238 }
239
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]240 SubStream i_a(graph);
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]241 i_a << ConvolutionLayer(
242 1U, 1U, a_filt,
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]243 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]244 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
245 PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]246 .set_name(param_path + "/Branch_0/Conv2d_0a_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]247 << BatchNormalizationLayer(
248 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
249 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
250 get_random_accessor(1.f, 1.f),
251 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]252 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]253 .set_name(param_path + "/Branch_0/Conv2d_0a_1x1/BatchNorm/batchnorm")
254 << 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]255
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]256 SubStream i_b(graph);
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]257 i_b << ConvolutionLayer(
258 1U, 1U, std::get<0>(b_filters),
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]259 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d" + conv_id0 + "1x1_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]260 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
261 PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]262 .set_name(param_path + "/Branch_1/Conv2d" + conv_id0 + "1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]263 << BatchNormalizationLayer(
264 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d" + conv_id0 + "1x1_BatchNorm_moving_mean.npy"),
265 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d" + conv_id0 + "1x1_BatchNorm_moving_variance.npy"),
266 get_random_accessor(1.f, 1.f),
267 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]268 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]269 .set_name(param_path + "/Branch_1/Conv2d" + conv_id0 + "1x1/BatchNorm/batchnorm")
270 << 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]271 << ConvolutionLayer(
272 5U, 5U, std::get<1>(b_filters),
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]273 get_weights_accessor(data_path, total_path + "Branch_1_Conv" + conv_id1 + "5x5_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]274 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
275 PadStrideInfo(1, 1, 2, 2))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]276 .set_name(param_path + "/Branch_1/Conv2d" + conv_id1 + "5x5/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]277 << BatchNormalizationLayer(
278 get_weights_accessor(data_path, total_path + "Branch_1_Conv" + conv_id1 + "5x5_BatchNorm_moving_mean.npy"),
279 get_weights_accessor(data_path, total_path + "Branch_1_Conv" + conv_id1 + "5x5_BatchNorm_moving_variance.npy"),
280 get_random_accessor(1.f, 1.f),
281 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]282 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]283 .set_name(param_path + "/Branch_1/Conv2d" + conv_id1 + "5x5/BatchNorm/batchnorm")
284 << 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]285
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]286 SubStream i_c(graph);
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]287 i_c << ConvolutionLayer(
288 1U, 1U, std::get<0>(c_filters),
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]289 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]290 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
291 PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]292 .set_name(param_path + "/Branch_2/Conv2d_0a_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]293 << BatchNormalizationLayer(
294 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
295 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
296 get_random_accessor(1.f, 1.f),
297 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]298 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]299 .set_name(param_path + "/Branch_2/Conv2d_0a_1x1/BatchNorm/batchnorm")
300 << 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]301 << ConvolutionLayer(
302 3U, 3U, std::get<1>(c_filters),
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]303 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_3x3_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]304 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
305 PadStrideInfo(1, 1, 1, 1))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]306 .set_name(param_path + "/Branch_2/Conv2d_0b_3x3/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]307 << BatchNormalizationLayer(
308 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_3x3_BatchNorm_moving_mean.npy"),
309 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_3x3_BatchNorm_moving_variance.npy"),
310 get_random_accessor(1.f, 1.f),
311 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]312 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]313 .set_name(param_path + "/Branch_2/Conv2d_0b_3x3/BatchNorm/batchnorm")
314 << 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]315 << ConvolutionLayer(
316 3U, 3U, std::get<2>(c_filters),
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]317 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_3x3_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]318 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
319 PadStrideInfo(1, 1, 1, 1))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]320 .set_name(param_path + "/Branch_2/Conv2d_0c_3x3/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]321 << BatchNormalizationLayer(
322 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_3x3_BatchNorm_moving_mean.npy"),
323 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_3x3_BatchNorm_moving_variance.npy"),
324 get_random_accessor(1.f, 1.f),
325 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]326 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]327 .set_name(param_path + "/Branch_2/Conv2d_0c_3x3/BatchNorm/batcnorm")
328 << 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]329
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]330 SubStream i_d(graph);
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]331 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]332 << ConvolutionLayer(
333 1U, 1U, d_filt,
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]334 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]335 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
336 PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]337 .set_name(param_path + "/Branch_3/Conv2d_0b_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]338 << BatchNormalizationLayer(
339 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_BatchNorm_moving_mean.npy"),
340 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_BatchNorm_moving_variance.npy"),
341 get_random_accessor(1.f, 1.f),
342 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]343 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]344 .set_name(param_path + "/Branch_3/Conv2d_0b_1x1/BatchNorm/batchnorm")
345 << 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]346
Georgios Pinitas427bbbf2018-08-28 13:32:02 +0100[diff] [blame]347 return ConcatLayer(std::move(i_a), std::move(i_b), std::move(i_c), std::move(i_d));
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]348 }
349
Georgios Pinitas427bbbf2018-08-28 13:32:02 +0100[diff] [blame]350 ConcatLayer get_inception_node_B(const std::string &data_path, std::string &&param_path, DataLayout weights_layout,
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]351 unsigned int a_filt,
352 std::tuple<unsigned int, unsigned int, unsigned int> b_filters)
353 {
354 std::string total_path = "/cnn_data/inceptionv3_model/" + param_path + "_";
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]355 SubStream i_a(graph);
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]356 i_a << ConvolutionLayer(
357 3U, 3U, a_filt,
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]358 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_1a_1x1_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]359 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
360 PadStrideInfo(2, 2, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]361 .set_name(param_path + "/Branch_0/Conv2d_1a_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]362 << BatchNormalizationLayer(
363 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_1a_1x1_BatchNorm_moving_mean.npy"),
364 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_1a_1x1_BatchNorm_moving_variance.npy"),
365 get_random_accessor(1.f, 1.f),
366 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]367 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]368 .set_name(param_path + "/Branch_0/Conv2d_1a_1x1/BatchNorm/batchnorm")
369 << 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]370
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]371 SubStream i_b(graph);
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]372 i_b << ConvolutionLayer(
373 1U, 1U, std::get<0>(b_filters),
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]374 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]375 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
376 PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]377 .set_name(param_path + "/Branch_1/Conv2d_0a_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]378 << BatchNormalizationLayer(
379 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
380 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
381 get_random_accessor(1.f, 1.f),
382 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]383 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]384 .set_name(param_path + "/Branch_1/Conv2d_0a_1x1/BatchNorm/batchnorm")
385 << 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]386 << ConvolutionLayer(
387 3U, 3U, std::get<1>(b_filters),
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]388 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_3x3_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]389 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
390 PadStrideInfo(1, 1, 1, 1))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]391 .set_name(param_path + "/Branch_1/Conv2d_0b_3x3/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]392 << BatchNormalizationLayer(
393 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_3x3_BatchNorm_moving_mean.npy"),
394 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_3x3_BatchNorm_moving_variance.npy"),
395 get_random_accessor(1.f, 1.f),
396 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]397 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]398 .set_name(param_path + "/Branch_1/Conv2d_0b_3x3/BatchNorm/batchnorm")
399 << 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]400 << ConvolutionLayer(
401 3U, 3U, std::get<2>(b_filters),
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]402 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_1a_1x1_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]403 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
404 PadStrideInfo(2, 2, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]405 .set_name(param_path + "/Branch_1/Conv2d_1a_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]406 << BatchNormalizationLayer(
407 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_1a_1x1_BatchNorm_moving_mean.npy"),
408 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_1a_1x1_BatchNorm_moving_variance.npy"),
409 get_random_accessor(1.f, 1.f),
410 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]411 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]412 .set_name(param_path + "/Branch_1/Conv2d_1a_1x1/BatchNorm/batchnorm")
413 << 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]414
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]415 SubStream i_c(graph);
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]416 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]417
Georgios Pinitas427bbbf2018-08-28 13:32:02 +0100[diff] [blame]418 return ConcatLayer(std::move(i_a), std::move(i_b), std::move(i_c));
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]419 }
420
Georgios Pinitas427bbbf2018-08-28 13:32:02 +0100[diff] [blame]421 ConcatLayer get_inception_node_C(const std::string &data_path, std::string &&param_path, DataLayout weights_layout,
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]422 unsigned int a_filt,
423 std::tuple<unsigned int, unsigned int, unsigned int> b_filters,
424 std::tuple<unsigned int, unsigned int, unsigned int, unsigned int, unsigned int> c_filters,
425 unsigned int d_filt)
426 {
427 std::string total_path = "/cnn_data/inceptionv3_model/" + param_path + "_";
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]428 SubStream i_a(graph);
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]429 i_a << ConvolutionLayer(
430 1U, 1U, a_filt,
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]431 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]432 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
433 PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]434 .set_name(param_path + "/Branch_0/Conv2d_0a_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]435 << BatchNormalizationLayer(
436 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
437 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
438 get_random_accessor(1.f, 1.f),
439 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]440 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]441 .set_name(param_path + "/Branch_0/Conv2d_0a_1x1/BatchNorm/batchnorm")
442 << 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]443
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]444 SubStream i_b(graph);
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]445 i_b << ConvolutionLayer(
446 1U, 1U, std::get<0>(b_filters),
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]447 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]448 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
449 PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]450 .set_name(param_path + "/Branch_1/Conv2d_0a_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]451 << BatchNormalizationLayer(
452 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
453 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
454 get_random_accessor(1.f, 1.f),
455 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]456 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]457 .set_name(param_path + "/Branch_1/Conv2d_0a_1x1/BatchNorm/batchnorm")
458 << 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]459 << ConvolutionLayer(
460 7U, 1U, std::get<1>(b_filters),
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]461 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x7_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]462 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
463 PadStrideInfo(1, 1, 3, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]464 .set_name(param_path + "/Branch_1/Conv2d_0b_1x7/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]465 << BatchNormalizationLayer(
466 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x7_BatchNorm_moving_mean.npy"),
467 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x7_BatchNorm_moving_variance.npy"),
468 get_random_accessor(1.f, 1.f),
469 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]470 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]471 .set_name(param_path + "/Branch_1/Conv2d_0b_1x7/BatchNorm/batchnorm")
472 << 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]473 << ConvolutionLayer(
474 1U, 7U, std::get<2>(b_filters),
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]475 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0c_7x1_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]476 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
477 PadStrideInfo(1, 1, 0, 3))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]478 .set_name(param_path + "/Branch_1/Conv2d_0c_7x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]479 << BatchNormalizationLayer(
480 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0c_7x1_BatchNorm_moving_mean.npy"),
481 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0c_7x1_BatchNorm_moving_variance.npy"),
482 get_random_accessor(1.f, 1.f),
483 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]484 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]485 .set_name(param_path + "/Branch_1/Conv2d_0c_7x1/BatchNorm/batchnorm")
486 << 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]487
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]488 SubStream i_c(graph);
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]489 i_c << ConvolutionLayer(
490 1U, 1U, std::get<0>(c_filters),
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]491 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]492 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
493 PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]494 .set_name(param_path + "/Branch_2/Conv2d_0a_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]495 << BatchNormalizationLayer(
496 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
497 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
498 get_random_accessor(1.f, 1.f),
499 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]500 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]501 .set_name(param_path + "/Branch_2/Conv2d_0a_1x1/BatchNorm/batchnorm")
502 << 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]503 << ConvolutionLayer(
504 1U, 7U, std::get<1>(c_filters),
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]505 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_7x1_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]506 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
507 PadStrideInfo(1, 1, 0, 3))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]508 .set_name(param_path + "/Branch_2/Conv2d_0b_7x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]509 << BatchNormalizationLayer(
510 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_7x1_BatchNorm_moving_mean.npy"),
511 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_7x1_BatchNorm_moving_variance.npy"),
512 get_random_accessor(1.f, 1.f),
513 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]514 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]515 .set_name(param_path + "/Branch_2/Conv2d_0b_7x1/BatchNorm/batchnorm")
516 << 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]517 << ConvolutionLayer(
518 7U, 1U, std::get<2>(c_filters),
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]519 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_1x7_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]520 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
521 PadStrideInfo(1, 1, 3, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]522 .set_name(param_path + "/Branch_2/Conv2d_0c_1x7/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]523 << BatchNormalizationLayer(
524 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_1x7_BatchNorm_moving_mean.npy"),
525 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_1x7_BatchNorm_moving_variance.npy"),
526 get_random_accessor(1.f, 1.f),
527 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]528 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]529 .set_name(param_path + "/Branch_2/Conv2d_0c_1x7/BatchNorm/batchnorm")
530 << 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]531 << ConvolutionLayer(
532 1U, 7U, std::get<3>(c_filters),
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]533 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0d_7x1_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]534 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
535 PadStrideInfo(1, 1, 0, 3))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]536 .set_name(param_path + "/Branch_2/Conv2d_0d_7x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]537 << BatchNormalizationLayer(
538 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0d_7x1_BatchNorm_moving_mean.npy"),
539 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0d_7x1_BatchNorm_moving_variance.npy"),
540 get_random_accessor(1.f, 1.f),
541 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]542 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]543 .set_name(param_path + "/Branch_2/Conv2d_0d_7x1/BatchNorm/batchnorm")
544 << 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]545 << ConvolutionLayer(
546 7U, 1U, std::get<4>(c_filters),
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]547 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0e_1x7_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]548 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
549 PadStrideInfo(1, 1, 3, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]550 .set_name(param_path + "/Branch_2/Conv2d_0e_1x7/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]551 << BatchNormalizationLayer(
552 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0e_1x7_BatchNorm_moving_mean.npy"),
553 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0e_1x7_BatchNorm_moving_variance.npy"),
554 get_random_accessor(1.f, 1.f),
555 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]556 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]557 .set_name(param_path + "/Branch_2/Conv2d_0e_1x7/BatchNorm/batchnorm")
558 << 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]559
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]560 SubStream i_d(graph);
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]561 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]562 << ConvolutionLayer(
563 1U, 1U, d_filt,
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]564 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]565 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
566 PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]567 .set_name(param_path + "/Branch_3/Conv2d_0b_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]568 << BatchNormalizationLayer(
569 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_BatchNorm_moving_mean.npy"),
570 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_BatchNorm_moving_variance.npy"),
571 get_random_accessor(1.f, 1.f),
572 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]573 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]574 .set_name(param_path + "/Branch_3/Conv2d_0b_1x1/BatchNorm/batchnorm")
575 << 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]576
Georgios Pinitas427bbbf2018-08-28 13:32:02 +0100[diff] [blame]577 return ConcatLayer(std::move(i_a), std::move(i_b), std::move(i_c), std::move(i_d));
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]578 }
579
Georgios Pinitas427bbbf2018-08-28 13:32:02 +0100[diff] [blame]580 ConcatLayer get_inception_node_D(const std::string &data_path, std::string &&param_path, DataLayout weights_layout,
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]581 std::tuple<unsigned int, unsigned int> a_filters,
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]582 std::tuple<unsigned int, unsigned int, unsigned int, unsigned int> b_filters)
583 {
584 std::string total_path = "/cnn_data/inceptionv3_model/" + param_path + "_";
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]585 SubStream i_a(graph);
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]586 i_a << ConvolutionLayer(
587 1U, 1U, std::get<0>(a_filters),
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]588 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]589 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
590 PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]591 .set_name(param_path + "/Branch_0/Conv2d_0a_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]592 << BatchNormalizationLayer(
593 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
594 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
595 get_random_accessor(1.f, 1.f),
596 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]597 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]598 .set_name(param_path + "/Branch_0/Conv2d_0a_1x1/BatchNorm/batchnorm")
599 << 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]600 << ConvolutionLayer(
601 3U, 3U, std::get<1>(a_filters),
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]602 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_1a_3x3_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]603 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
604 PadStrideInfo(2, 2, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]605 .set_name(param_path + "/Branch_0/Conv2d_1a_3x3/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]606 << BatchNormalizationLayer(
607 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_1a_3x3_BatchNorm_moving_mean.npy"),
608 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_1a_3x3_BatchNorm_moving_variance.npy"),
609 get_random_accessor(1.f, 1.f),
610 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]611 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]612 .set_name(param_path + "/Branch_0/Conv2d_1a_3x3/BatchNorm/batchnorm")
613 << 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]614
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]615 SubStream i_b(graph);
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]616 i_b << ConvolutionLayer(
617 1U, 1U, std::get<0>(b_filters),
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]618 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]619 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
620 PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]621 .set_name(param_path + "/Branch_1/Conv2d_0a_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]622 << BatchNormalizationLayer(
623 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
624 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
625 get_random_accessor(1.f, 1.f),
626 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]627 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]628 .set_name(param_path + "/Branch_1/Conv2d_0a_1x1/BatchNorm/batchnorm")
629 << 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]630 << ConvolutionLayer(
631 7U, 1U, std::get<1>(b_filters),
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]632 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x7_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]633 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
634 PadStrideInfo(1, 1, 3, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]635 .set_name(param_path + "/Branch_1/Conv2d_0b_1x7/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]636 << BatchNormalizationLayer(
637 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x7_BatchNorm_moving_mean.npy"),
638 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x7_BatchNorm_moving_variance.npy"),
639 get_random_accessor(1.f, 1.f),
640 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]641 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]642 .set_name(param_path + "/Branch_1/Conv2d_0b_1x7/BatchNorm/batchnorm")
643 << 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]644 << ConvolutionLayer(
645 1U, 7U, std::get<2>(b_filters),
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]646 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0c_7x1_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]647 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
648 PadStrideInfo(1, 1, 0, 3))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]649 .set_name(param_path + "/Branch_1/Conv2d_0c_7x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]650 << BatchNormalizationLayer(
651 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0c_7x1_BatchNorm_moving_mean.npy"),
652 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0c_7x1_BatchNorm_moving_variance.npy"),
653 get_random_accessor(1.f, 1.f),
654 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]655 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]656 .set_name(param_path + "/Branch_1/Conv2d_0c_7x1/BatchNorm/batchnorm")
657 << 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]658 << ConvolutionLayer(
659 3U, 3U, std::get<3>(b_filters),
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]660 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_1a_3x3_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]661 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
662 PadStrideInfo(2, 2, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]663 .set_name(param_path + "/Branch_1/Conv2d_1a_3x3/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]664 << BatchNormalizationLayer(
665 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_1a_3x3_BatchNorm_moving_mean.npy"),
666 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_1a_3x3_BatchNorm_moving_variance.npy"),
667 get_random_accessor(1.f, 1.f),
668 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]669 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]670 .set_name(param_path + "/Branch_1/Conv2d_1a_3x3/BatchNorm/batchnorm")
671 << 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]672
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]673 SubStream i_c(graph);
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]674 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]675
Georgios Pinitas427bbbf2018-08-28 13:32:02 +0100[diff] [blame]676 return ConcatLayer(std::move(i_a), std::move(i_b), std::move(i_c));
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]677 }
678
Georgios Pinitas427bbbf2018-08-28 13:32:02 +0100[diff] [blame]679 ConcatLayer get_inception_node_E(const std::string &data_path, std::string &&param_path, DataLayout weights_layout,
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]680 unsigned int a_filt,
681 std::tuple<unsigned int, unsigned int, unsigned int> b_filters,
682 std::tuple<unsigned int, unsigned int, unsigned int, unsigned int> c_filters,
683 unsigned int d_filt,
684 bool is_name_different = false)
685 {
686 // This is due to a naming issue in the tf model
687 std::string conv_id = "_0b_";
688 if(is_name_different)
689 {
690 conv_id = "_0c_";
691 }
692
693 std::string total_path = "/cnn_data/inceptionv3_model/" + param_path + "_";
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]694 SubStream i_a(graph);
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]695 i_a << ConvolutionLayer(
696 1U, 1U, a_filt,
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]697 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]698 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
699 PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]700 .set_name(param_path + "/Branch_0/Conv2d_0a_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]701 << BatchNormalizationLayer(
702 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
703 get_weights_accessor(data_path, total_path + "Branch_0_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
704 get_random_accessor(1.f, 1.f),
705 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]706 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]707 .set_name(param_path + "/Branch_0/Conv2d_0a_1x1/BatchNorm/batchnorm")
708 << 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]709
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]710 SubStream i_b(graph);
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]711 i_b << ConvolutionLayer(
712 1U, 1U, std::get<0>(b_filters),
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]713 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]714 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
715 PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]716 .set_name(param_path + "/Branch_1/Conv2d_0a_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]717 << BatchNormalizationLayer(
718 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
719 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
720 get_random_accessor(1.f, 1.f),
721 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]722 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]723 .set_name(param_path + "/Branch_1/Conv2d_0a_1x1/BatchNorm/batchnorm")
724 << 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]725
Georgios Pinitas772e17f2018-07-13 12:25:33 +0100[diff] [blame]726 SubStream i_b1(i_b);
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]727 i_b1 << ConvolutionLayer(
728 3U, 1U, std::get<1>(b_filters),
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]729 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x3_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]730 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
731 PadStrideInfo(1, 1, 1, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]732 .set_name(param_path + "/Branch_1/Conv2d_0b_1x3/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]733 << BatchNormalizationLayer(
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]734 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x3_BatchNorm_moving_mean.npy"),
735 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]736 get_random_accessor(1.f, 1.f),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]737 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d_0b_1x3_BatchNorm_beta.npy"),
738 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]739 .set_name(param_path + "/Branch_1/Conv2d_0b_1x3/BatchNorm/batchnorm")
740 << 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]741
Georgios Pinitas772e17f2018-07-13 12:25:33 +0100[diff] [blame]742 SubStream i_b2(i_b);
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]743 i_b2 << ConvolutionLayer(
744 1U, 3U, std::get<2>(b_filters),
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]745 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d" + conv_id + "3x1_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]746 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
747 PadStrideInfo(1, 1, 0, 1))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]748 .set_name(param_path + "/Branch_1/Conv2d" + conv_id + "3x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]749 << BatchNormalizationLayer(
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]750 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d" + conv_id + "3x1_BatchNorm_moving_mean.npy"),
751 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]752 get_random_accessor(1.f, 1.f),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]753 get_weights_accessor(data_path, total_path + "Branch_1_Conv2d" + conv_id + "3x1_BatchNorm_beta.npy"),
754 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]755 .set_name(param_path + "/Branch_1/Conv2d" + conv_id + "3x1/BatchNorm/batchnorm")
756 << 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]757
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]758 // Merge b1 and b2
Georgios Pinitas427bbbf2018-08-28 13:32:02 +0100[diff] [blame]759 i_b << ConcatLayer(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]760
761 SubStream i_c(graph);
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]762 i_c << ConvolutionLayer(
763 1U, 1U, std::get<0>(c_filters),
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]764 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]765 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
766 PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]767 .set_name(param_path + "/Branch_2/Conv2d_0a_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]768 << BatchNormalizationLayer(
769 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_BatchNorm_moving_mean.npy"),
770 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0a_1x1_BatchNorm_moving_variance.npy"),
771 get_random_accessor(1.f, 1.f),
772 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]773 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]774 .set_name(param_path + "/Branch_2/Conv2d_0a_1x1/BatchNorm/batchnorm")
775 << 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]776 << ConvolutionLayer(
777 3U, 3U, std::get<1>(c_filters),
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]778 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_3x3_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]779 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
780 PadStrideInfo(1, 1, 1, 1))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]781 .set_name(param_path + "/Branch_2/Conv2d_0b_3x3/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]782 << BatchNormalizationLayer(
783 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_3x3_BatchNorm_moving_mean.npy"),
784 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0b_3x3_BatchNorm_moving_variance.npy"),
785 get_random_accessor(1.f, 1.f),
786 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]787 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]788 .set_name(param_path + "/Branch_2/Conv2d_0b_3x3/BatchNorm/batchnorm")
789 << 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]790
Georgios Pinitas772e17f2018-07-13 12:25:33 +0100[diff] [blame]791 SubStream i_c1(i_c);
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]792 i_c1 << ConvolutionLayer(
793 3U, 1U, std::get<2>(c_filters),
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]794 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_1x3_weights.npy", weights_layout),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]795 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
796 PadStrideInfo(1, 1, 1, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]797 .set_name(param_path + "/Branch_2/Conv2d_0c_1x3/convolution")
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]798 << BatchNormalizationLayer(
799 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_1x3_BatchNorm_moving_mean.npy"),
800 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_1x3_BatchNorm_moving_variance.npy"),
801 get_random_accessor(1.f, 1.f),
802 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0c_1x3_BatchNorm_beta.npy"),
803 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]804 .set_name(param_path + "/Branch_2/Conv2d_0c_1x3/BatchNorm/batchnorm")
805 << 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]806
Georgios Pinitas772e17f2018-07-13 12:25:33 +0100[diff] [blame]807 SubStream i_c2(i_c);
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]808 i_c2 << ConvolutionLayer(
809 1U, 3U, std::get<3>(c_filters),
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]810 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0d_3x1_weights.npy", weights_layout),
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]811 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
812 PadStrideInfo(1, 1, 0, 1))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]813 .set_name(param_path + "/Branch_2/Conv2d_0d_3x1/convolution")
Georgios Pinitasd8734b52017-12-22 15:27:52 +0000[diff] [blame]814 << BatchNormalizationLayer(
815 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0d_3x1_BatchNorm_moving_mean.npy"),
816 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0d_3x1_BatchNorm_moving_variance.npy"),
817 get_random_accessor(1.f, 1.f),
818 get_weights_accessor(data_path, total_path + "Branch_2_Conv2d_0d_3x1_BatchNorm_beta.npy"),
819 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]820 .set_name(param_path + "/Branch_2/Conv2d_0d_3x1/BatchNorm/batchnorm")
821 << 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]822
823 // Merge i_c1 and i_c2
Georgios Pinitas427bbbf2018-08-28 13:32:02 +0100[diff] [blame]824 i_c << ConcatLayer(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]825
826 SubStream i_d(graph);
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]827 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]828 << ConvolutionLayer(
829 1U, 1U, d_filt,
Georgios Pinitase2220552018-07-20 13:23:44 +0100[diff] [blame]830 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_weights.npy", weights_layout),
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]831 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
832 PadStrideInfo(1, 1, 0, 0))
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]833 .set_name(param_path + "/Branch_3/Conv2d_0b_1x1/convolution")
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]834 << BatchNormalizationLayer(
835 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_BatchNorm_moving_mean.npy"),
836 get_weights_accessor(data_path, total_path + "Branch_3_Conv2d_0b_1x1_BatchNorm_moving_variance.npy"),
837 get_random_accessor(1.f, 1.f),
838 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]839 0.001f)
Georgios Pinitas130986a2018-05-14 19:25:37 +0100[diff] [blame]840 .set_name(param_path + "/Branch_3/Conv2d_0b_1x1/BatchNorm/batchnorm")
841 << 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]842
Georgios Pinitas427bbbf2018-08-28 13:32:02 +0100[diff] [blame]843 return ConcatLayer(std::move(i_a), std::move(i_b), std::move(i_c), std::move(i_d));
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]844 }
845};
846
847/** Main program for Inception V3
848 *
Georgios Pinitasbdbbbe82018-11-07 16:06:47 +0000[diff] [blame]849 * Model is based on:
850 * https://arxiv.org/abs/1512.00567
851 * "Rethinking the Inception Architecture for Computer Vision"
852 * Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe, Jonathon Shlens, Zbigniew Wojna
853 *
Georgios Pinitas9f28b392018-07-18 20:01:53 +0100[diff] [blame]854 * @note To list all the possible arguments execute the binary appended with the --help option
855 *
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]856 * @param[in] argc Number of arguments
Georgios Pinitas12be7ab2018-07-03 12:06:23 +0100[diff] [blame]857 * @param[in] argv Arguments
Georgios Pinitas652bde52018-01-10 15:33:28 +0000[diff] [blame]858 */
859int main(int argc, char **argv)
860{
861 return arm_compute::utils::run_example<InceptionV3Example>(argc, argv);
862}