Gitiles
Code Review Sign In
review.mlplatform.org / ml / ComputeLibrary / 04a8f8c4994f1c32b3f16a832c0e6f2599364c02 / . / examples / graph_mobilenet.cpp
blob: 2b2da9e517c507da193cf6f6095e8312272b66d5 [file] [log] [blame]
Georgios Pinitas236bfe72017-11-23 15:59:55 +0000[diff] [blame]1/*
2 * Copyright (c) 2017 ARM Limited.
3 *
4 * SPDX-License-Identifier: MIT
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to
8 * deal in the Software without restriction, including without limitation the
9 * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
10 * sell copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
12 *
13 * The above copyright notice and this permission notice shall be included in all
14 * copies or substantial portions of the Software.
15 *
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
19 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
22 * SOFTWARE.
23 */
24
25#include "arm_compute/graph/Graph.h"
26#include "arm_compute/graph/Nodes.h"
27#include "support/ToolchainSupport.h"
28#include "utils/GraphUtils.h"
29#include "utils/Utils.h"
30
31#include <cstdlib>
32
33using namespace arm_compute::graph;
34using namespace arm_compute::graph_utils;
35
36BranchLayer get_dwsc_node(const std::string &data_path, std::string &&param_path,
37 unsigned int conv_filt,
38 PadStrideInfo dwc_pad_stride_info, PadStrideInfo conv_pad_stride_info)
39{
40 std::string total_path = "/cnn_data/mobilenet_v1_model/" + param_path + "_";
41 SubGraph sg;
42 sg << DepthwiseConvolutionLayer(
43 3U, 3U,
44 get_weights_accessor(data_path, total_path + "depthwise_depthwise_weights.npy"),
45 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
46 dwc_pad_stride_info,
47 true)
48 << BatchNormalizationLayer(
49 get_weights_accessor(data_path, total_path + "depthwise_BatchNorm_moving_mean.npy"),
50 get_weights_accessor(data_path, total_path + "depthwise_BatchNorm_moving_variance.npy"),
51 get_weights_accessor(data_path, total_path + "depthwise_BatchNorm_beta.npy"),
52 get_weights_accessor(data_path, total_path + "depthwise_BatchNorm_gamma.npy"),
53 0.001f)
54 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::BOUNDED_RELU, 6.f))
55 << ConvolutionLayer(
56 1U, 1U, conv_filt,
57 get_weights_accessor(data_path, total_path + "pointwise_weights.npy"),
58 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
59 conv_pad_stride_info)
60 << BatchNormalizationLayer(
61 get_weights_accessor(data_path, total_path + "pointwise_BatchNorm_moving_mean.npy"),
62 get_weights_accessor(data_path, total_path + "pointwise_BatchNorm_moving_variance.npy"),
63 get_weights_accessor(data_path, total_path + "pointwise_BatchNorm_beta.npy"),
64 get_weights_accessor(data_path, total_path + "pointwise_BatchNorm_gamma.npy"),
65 0.001f)
66 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::BOUNDED_RELU, 6.f));
67
68 return BranchLayer(std::move(sg));
69}
70
71/** Example demonstrating how to implement MobileNet's network using the Compute Library's graph API
72 *
73 * @param[in] argc Number of arguments
74 * @param[in] argv Arguments ( [optional] Path to the weights folder, [optional] image, [optional] labels )
75 */
76void main_graph_mobilenet(int argc, const char **argv)
77{
78 std::string data_path; /* Path to the trainable data */
79 std::string image; /* Image data */
80 std::string label; /* Label data */
81
82 constexpr float mean_r = 122.68f; /* Mean value to subtract from red channel */
83 constexpr float mean_g = 116.67f; /* Mean value to subtract from green channel */
84 constexpr float mean_b = 104.01f; /* Mean value to subtract from blue channel */
85
86 // Parse arguments
87 if(argc < 2)
88 {
89 // Print help
90 std::cout << "Usage: " << argv[0] << " [path_to_data] [image] [labels]\n\n";
91 std::cout << "No data folder provided: using random values\n\n";
92 }
93 else if(argc == 2)
94 {
95 data_path = argv[1];
96 std::cout << "Usage: " << argv[0] << " " << argv[1] << " [image] [labels]\n\n";
97 std::cout << "No image provided: using random values\n\n";
98 }
99 else if(argc == 3)
100 {
101 data_path = argv[1];
102 image = argv[2];
103 std::cout << "Usage: " << argv[0] << " " << argv[1] << " " << argv[2] << " [labels]\n\n";
104 std::cout << "No text file with labels provided: skipping output accessor\n\n";
105 }
106 else
107 {
108 data_path = argv[1];
109 image = argv[2];
110 label = argv[3];
111 }
112
113 // Check if OpenCL is available and initialize the scheduler
114 TargetHint hint = TargetHint::NEON;
115 if(Graph::opencl_is_available())
116 {
117 hint = TargetHint::OPENCL;
118 }
119
120 Graph graph;
121 graph << hint
122 << Tensor(TensorInfo(TensorShape(224U, 224U, 3U, 1U), 1, DataType::F32),
123 get_input_accessor(image, mean_r, mean_g, mean_b))
124 << ConvolutionLayer(
125 3U, 3U, 32U,
126 get_weights_accessor(data_path, "/cnn_data/mobilenet_v1_model/Conv2d_0_weights.npy"),
127 std::unique_ptr<arm_compute::graph::ITensorAccessor>(nullptr),
128 PadStrideInfo(2, 2, 0, 1, 0, 1, DimensionRoundingType::FLOOR))
129 << BatchNormalizationLayer(
130 get_weights_accessor(data_path, "/cnn_data/mobilenet_v1_model/Conv2d_0_BatchNorm_moving_mean.npy"),
131 get_weights_accessor(data_path, "/cnn_data/mobilenet_v1_model/Conv2d_0_BatchNorm_moving_variance.npy"),
132 get_weights_accessor(data_path, "/cnn_data/mobilenet_v1_model/Conv2d_0_BatchNorm_beta.npy"),
133 get_weights_accessor(data_path, "/cnn_data/mobilenet_v1_model/Conv2d_0_BatchNorm_gamma.npy"),
134 0.001f)
135 << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::BOUNDED_RELU, 6.f))
136 << get_dwsc_node(data_path, "Conv2d_1", 64, PadStrideInfo(1, 1, 1, 1), PadStrideInfo(1, 1, 0, 0))
137 << get_dwsc_node(data_path, "Conv2d_2", 128, PadStrideInfo(2, 2, 0, 1, 0, 1, DimensionRoundingType::FLOOR), PadStrideInfo(1, 1, 0, 0))
138 << get_dwsc_node(data_path, "Conv2d_3", 128, PadStrideInfo(1, 1, 1, 1, 1, 1, DimensionRoundingType::FLOOR), PadStrideInfo(1, 1, 0, 0))
139 << get_dwsc_node(data_path, "Conv2d_4", 256, PadStrideInfo(2, 2, 0, 1, 0, 1, DimensionRoundingType::FLOOR), PadStrideInfo(1, 1, 0, 0))
140 << get_dwsc_node(data_path, "Conv2d_5", 256, PadStrideInfo(1, 1, 1, 1, 1, 1, DimensionRoundingType::FLOOR), PadStrideInfo(1, 1, 0, 0))
141 << get_dwsc_node(data_path, "Conv2d_6", 512, PadStrideInfo(2, 2, 0, 1, 0, 1, DimensionRoundingType::FLOOR), PadStrideInfo(1, 1, 0, 0))
142 << get_dwsc_node(data_path, "Conv2d_7", 512, PadStrideInfo(1, 1, 1, 1, 1, 1, DimensionRoundingType::FLOOR), PadStrideInfo(1, 1, 0, 0))
143 << get_dwsc_node(data_path, "Conv2d_8", 512, PadStrideInfo(1, 1, 1, 1, 1, 1, DimensionRoundingType::FLOOR), PadStrideInfo(1, 1, 0, 0))
144 << get_dwsc_node(data_path, "Conv2d_9", 512, PadStrideInfo(1, 1, 1, 1, 1, 1, DimensionRoundingType::FLOOR), PadStrideInfo(1, 1, 0, 0))
145 << get_dwsc_node(data_path, "Conv2d_10", 512, PadStrideInfo(1, 1, 1, 1, 1, 1, DimensionRoundingType::FLOOR), PadStrideInfo(1, 1, 0, 0))
146 << get_dwsc_node(data_path, "Conv2d_11", 512, PadStrideInfo(1, 1, 1, 1, 1, 1, DimensionRoundingType::FLOOR), PadStrideInfo(1, 1, 0, 0))
147 << get_dwsc_node(data_path, "Conv2d_12", 1024, PadStrideInfo(2, 2, 0, 1, 0, 1, DimensionRoundingType::FLOOR), PadStrideInfo(1, 1, 0, 0))
148 << get_dwsc_node(data_path, "Conv2d_13", 1024, PadStrideInfo(1, 1, 1, 1, 1, 1, DimensionRoundingType::FLOOR), PadStrideInfo(1, 1, 0, 0))
149 << PoolingLayer(PoolingLayerInfo(PoolingType::AVG))
150 << ConvolutionLayer(
151 1U, 1U, 1001U,
152 get_weights_accessor(data_path, "/cnn_data/mobilenet_v1_model/Logits_Conv2d_1c_1x1_weights.npy"),
153 get_weights_accessor(data_path, "/cnn_data/mobilenet_v1_model/Logits_Conv2d_1c_1x1_biases.npy"),
154 PadStrideInfo(1, 1, 0, 0))
155 << ReshapeLayer(TensorShape(1001U))
156 << SoftmaxLayer()
157 << Tensor(get_output_accessor(label, 5));
158
159 graph.run();
160}
161
162/** Main program for MobileNetV1
163 *
164 * @param[in] argc Number of arguments
165 * @param[in] argv Arguments ( [optional] Path to the weights folder, [optional] image, [optional] labels )
166 */
167int main(int argc, const char **argv)
168{
169 return arm_compute::utils::run_example(argc, argv, main_graph_mobilenet);
170}