diff options
-rw-r--r-- | examples/graph_alexnet.cpp | 98 | ||||
-rw-r--r-- | examples/graph_googlenet.cpp | 96 | ||||
-rw-r--r-- | utils/GraphUtils.cpp | 3 | ||||
-rw-r--r-- | utils/GraphUtils.h | 66 | ||||
-rw-r--r-- | utils/Utils.h | 14 |
5 files changed, 143 insertions, 134 deletions
diff --git a/examples/graph_alexnet.cpp b/examples/graph_alexnet.cpp index 1d041997e3..b2a5be647f 100644 --- a/examples/graph_alexnet.cpp +++ b/examples/graph_alexnet.cpp @@ -42,72 +42,6 @@ using namespace arm_compute::graph; using namespace arm_compute::graph_utils; using namespace arm_compute::logging; -/** Generates appropriate accessor according to the specified path - * - * @note If path is empty will generate a DummyAccessor else will generate a NumPyBinLoader - * - * @param[in] path Path to the data files - * @param[in] data_file Relative path to the data files from path - * - * @return An appropriate tensor accessor - */ -std::unique_ptr<ITensorAccessor> get_accessor(const std::string &path, const std::string &data_file) -{ - if(path.empty()) - { - return arm_compute::support::cpp14::make_unique<DummyAccessor>(); - } - else - { - return arm_compute::support::cpp14::make_unique<NumPyBinLoader>(path + data_file); - } -} - -/** Generates appropriate input accessor according to the specified ppm_path - * - * @note If ppm_path is empty will generate a DummyAccessor else will generate a PPMAccessor - * - * @param[in] ppm_path Path to PPM file - * @param[in] mean_r Red mean value to be subtracted from red channel - * @param[in] mean_g Green mean value to be subtracted from green channel - * @param[in] mean_b Blue mean value to be subtracted from blue channel - * - * @return An appropriate tensor accessor - */ -std::unique_ptr<ITensorAccessor> get_input_accessor(const std::string &ppm_path, float mean_r, float mean_g, float mean_b) -{ - if(ppm_path.empty()) - { - return arm_compute::support::cpp14::make_unique<DummyAccessor>(); - } - else - { - return arm_compute::support::cpp14::make_unique<PPMAccessor>(ppm_path, true, mean_r, mean_g, mean_b); - } -} - -/** Generates appropriate output accessor according to the specified labels_path - * - * @note If labels_path is empty will generate a DummyAccessor else will generate a TopNPredictionsAccessor - * - * @param[in] labels_path Path to labels text file - * @param[in] top_n (Optional) Number of output classes to print - * @param[out] output_stream (Optional) Output stream - * - * @return An appropriate tensor accessor - */ -std::unique_ptr<ITensorAccessor> get_output_accessor(const std::string &labels_path, size_t top_n = 5, std::ostream &output_stream = std::cout) -{ - if(labels_path.empty()) - { - return arm_compute::support::cpp14::make_unique<DummyAccessor>(); - } - else - { - return arm_compute::support::cpp14::make_unique<TopNPredictionsAccessor>(labels_path, top_n, output_stream); - } -} - /** Example demonstrating how to implement AlexNet's network using the Compute Library's graph API * * @param[in] argc Number of arguments @@ -166,8 +100,8 @@ void main_graph_alexnet(int argc, const char **argv) // Layer 1 << ConvolutionLayer( 11U, 11U, 96U, - get_accessor(data_path, "/cnn_data/alexnet_model/conv1_w.npy"), - get_accessor(data_path, "/cnn_data/alexnet_model/conv1_b.npy"), + get_weights_accessor(data_path, "/cnn_data/alexnet_model/conv1_w.npy"), + get_weights_accessor(data_path, "/cnn_data/alexnet_model/conv1_b.npy"), PadStrideInfo(4, 4, 0, 0)) << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)) << NormalizationLayer(NormalizationLayerInfo(NormType::CROSS_MAP, 5, 0.0001f, 0.75f)) @@ -176,8 +110,8 @@ void main_graph_alexnet(int argc, const char **argv) << ConvolutionMethodHint::DIRECT << ConvolutionLayer( 5U, 5U, 256U, - get_accessor(data_path, "/cnn_data/alexnet_model/conv2_w.npy"), - get_accessor(data_path, "/cnn_data/alexnet_model/conv2_b.npy"), + get_weights_accessor(data_path, "/cnn_data/alexnet_model/conv2_w.npy"), + get_weights_accessor(data_path, "/cnn_data/alexnet_model/conv2_b.npy"), PadStrideInfo(1, 1, 2, 2), 2) << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)) << NormalizationLayer(NormalizationLayerInfo(NormType::CROSS_MAP, 5, 0.0001f, 0.75f)) @@ -185,42 +119,42 @@ void main_graph_alexnet(int argc, const char **argv) // Layer 3 << ConvolutionLayer( 3U, 3U, 384U, - get_accessor(data_path, "/cnn_data/alexnet_model/conv3_w.npy"), - get_accessor(data_path, "/cnn_data/alexnet_model/conv3_b.npy"), + get_weights_accessor(data_path, "/cnn_data/alexnet_model/conv3_w.npy"), + get_weights_accessor(data_path, "/cnn_data/alexnet_model/conv3_b.npy"), PadStrideInfo(1, 1, 1, 1)) << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)) // Layer 4 << ConvolutionLayer( 3U, 3U, 384U, - get_accessor(data_path, "/cnn_data/alexnet_model/conv4_w.npy"), - get_accessor(data_path, "/cnn_data/alexnet_model/conv4_b.npy"), + get_weights_accessor(data_path, "/cnn_data/alexnet_model/conv4_w.npy"), + get_weights_accessor(data_path, "/cnn_data/alexnet_model/conv4_b.npy"), PadStrideInfo(1, 1, 1, 1), 2) << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)) // Layer 5 << ConvolutionLayer( 3U, 3U, 256U, - get_accessor(data_path, "/cnn_data/alexnet_model/conv5_w.npy"), - get_accessor(data_path, "/cnn_data/alexnet_model/conv5_b.npy"), + get_weights_accessor(data_path, "/cnn_data/alexnet_model/conv5_w.npy"), + get_weights_accessor(data_path, "/cnn_data/alexnet_model/conv5_b.npy"), PadStrideInfo(1, 1, 1, 1), 2) << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)) << PoolingLayer(PoolingLayerInfo(PoolingType::MAX, 3, PadStrideInfo(2, 2, 0, 0))) // Layer 6 << FullyConnectedLayer( 4096U, - get_accessor(data_path, "/cnn_data/alexnet_model/fc6_w.npy"), - get_accessor(data_path, "/cnn_data/alexnet_model/fc6_b.npy")) + get_weights_accessor(data_path, "/cnn_data/alexnet_model/fc6_w.npy"), + get_weights_accessor(data_path, "/cnn_data/alexnet_model/fc6_b.npy")) << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)) // Layer 7 << FullyConnectedLayer( 4096U, - get_accessor(data_path, "/cnn_data/alexnet_model/fc7_w.npy"), - get_accessor(data_path, "/cnn_data/alexnet_model/fc7_b.npy")) + get_weights_accessor(data_path, "/cnn_data/alexnet_model/fc7_w.npy"), + get_weights_accessor(data_path, "/cnn_data/alexnet_model/fc7_b.npy")) << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)) // Layer 8 << FullyConnectedLayer( 1000U, - get_accessor(data_path, "/cnn_data/alexnet_model/fc8_w.npy"), - get_accessor(data_path, "/cnn_data/alexnet_model/fc8_b.npy")) + get_weights_accessor(data_path, "/cnn_data/alexnet_model/fc8_w.npy"), + get_weights_accessor(data_path, "/cnn_data/alexnet_model/fc8_b.npy")) // Softmax << SoftmaxLayer() << Tensor(get_output_accessor(label, 5)); diff --git a/examples/graph_googlenet.cpp b/examples/graph_googlenet.cpp index 0e82c1e85d..354a2a39e4 100644 --- a/examples/graph_googlenet.cpp +++ b/examples/graph_googlenet.cpp @@ -42,27 +42,6 @@ using namespace arm_compute::graph; using namespace arm_compute::graph_utils; -/** Generates appropriate accessor according to the specified path - * - * @note If path is empty will generate a DummyAccessor else will generate a NumPyBinLoader - * - * @param path Path to the data files - * @param data_file Relative path to the data files from path - * - * @return An appropriate tensor accessor - */ -std::unique_ptr<ITensorAccessor> get_accessor(const std::string &path, const std::string &data_file) -{ - if(path.empty()) - { - return arm_compute::support::cpp14::make_unique<DummyAccessor>(); - } - else - { - return arm_compute::support::cpp14::make_unique<NumPyBinLoader>(path + data_file); - } -} - BranchLayer get_inception_node(const std::string &data_path, std::string &¶m_path, unsigned int a_filt, std::tuple<unsigned int, unsigned int> b_filters, @@ -73,36 +52,36 @@ BranchLayer get_inception_node(const std::string &data_path, std::string &¶m SubGraph i_a; i_a << ConvolutionLayer( 1U, 1U, a_filt, - get_accessor(data_path, total_path + "1x1_w.npy"), - get_accessor(data_path, total_path + "1x1_b.npy"), + get_weights_accessor(data_path, total_path + "1x1_w.npy"), + get_weights_accessor(data_path, total_path + "1x1_b.npy"), PadStrideInfo(1, 1, 0, 0)) << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)); SubGraph i_b; i_b << ConvolutionLayer( 1U, 1U, std::get<0>(b_filters), - get_accessor(data_path, total_path + "3x3_reduce_w.npy"), - get_accessor(data_path, total_path + "3x3_reduce_b.npy"), + get_weights_accessor(data_path, total_path + "3x3_reduce_w.npy"), + get_weights_accessor(data_path, total_path + "3x3_reduce_b.npy"), PadStrideInfo(1, 1, 0, 0)) << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)) << ConvolutionLayer( 3U, 3U, std::get<1>(b_filters), - get_accessor(data_path, total_path + "3x3_w.npy"), - get_accessor(data_path, total_path + "3x3_b.npy"), + get_weights_accessor(data_path, total_path + "3x3_w.npy"), + get_weights_accessor(data_path, total_path + "3x3_b.npy"), PadStrideInfo(1, 1, 1, 1)) << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)); SubGraph i_c; i_c << ConvolutionLayer( 1U, 1U, std::get<0>(c_filters), - get_accessor(data_path, total_path + "5x5_reduce_w.npy"), - get_accessor(data_path, total_path + "5x5_reduce_b.npy"), + get_weights_accessor(data_path, total_path + "5x5_reduce_w.npy"), + get_weights_accessor(data_path, total_path + "5x5_reduce_b.npy"), PadStrideInfo(1, 1, 0, 0)) << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)) << ConvolutionLayer( 5U, 5U, std::get<1>(c_filters), - get_accessor(data_path, total_path + "5x5_w.npy"), - get_accessor(data_path, total_path + "5x5_b.npy"), + get_weights_accessor(data_path, total_path + "5x5_w.npy"), + get_weights_accessor(data_path, total_path + "5x5_b.npy"), PadStrideInfo(1, 1, 2, 2)) << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)); @@ -110,8 +89,8 @@ BranchLayer get_inception_node(const std::string &data_path, std::string &¶m i_d << PoolingLayer(PoolingLayerInfo(PoolingType::MAX, 3, PadStrideInfo(1, 1, 1, 1, DimensionRoundingType::CEIL))) << ConvolutionLayer( 1U, 1U, d_filt, - get_accessor(data_path, total_path + "pool_proj_w.npy"), - get_accessor(data_path, total_path + "pool_proj_b.npy"), + get_weights_accessor(data_path, total_path + "pool_proj_w.npy"), + get_weights_accessor(data_path, total_path + "pool_proj_b.npy"), PadStrideInfo(1, 1, 0, 0)) << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)); @@ -121,32 +100,44 @@ BranchLayer get_inception_node(const std::string &data_path, std::string &¶m /** Example demonstrating how to implement Googlenet's network using the Compute Library's graph API * * @param[in] argc Number of arguments - * @param[in] argv Arguments ( [optional] Path to the weights folder, [optional] batches ) + * @param[in] argv Arguments ( [optional] Path to the weights folder, [optional] image, [optional] labels ) */ void main_graph_googlenet(int argc, const char **argv) { - std::string data_path; /** Path to the trainable data */ - unsigned int batches = 4; /** Number of batches */ + std::string data_path; /* Path to the trainable data */ + std::string image; /* Image data */ + std::string label; /* Label data */ + + constexpr float mean_r = 122.68f; /* Mean value to subtract from red channel */ + constexpr float mean_g = 116.67f; /* Mean value to subtract from green channel */ + constexpr float mean_b = 104.01f; /* Mean value to subtract from blue channel */ // Parse arguments if(argc < 2) { // Print help - std::cout << "Usage: " << argv[0] << " [path_to_data] [batches]\n\n"; + std::cout << "Usage: " << argv[0] << " [path_to_data] [image] [labels]\n\n"; std::cout << "No data folder provided: using random values\n\n"; } else if(argc == 2) { //Do something with argv[1] data_path = argv[1]; - std::cout << "Usage: " << argv[0] << " [path_to_data] [batches]\n\n"; - std::cout << "No number of batches where specified, thus will use the default : " << batches << "\n\n"; + std::cout << "Usage: " << argv[0] << " " << argv[1] << " [image] [labels]\n\n"; + std::cout << "No image provided: using random values\n"; + } + else if(argc == 3) + { + data_path = argv[1]; + image = argv[2]; + std::cout << "Usage: " << argv[0] << " " << argv[1] << " " << argv[2] << " [labels]\n\n"; + std::cout << "No text file with labels provided: skipping output accessor\n"; } else { - //Do something with argv[1] and argv[2] data_path = argv[1]; - batches = std::strtol(argv[2], nullptr, 0); + image = argv[2]; + label = argv[3]; } // Check if OpenCL is available and initialize the scheduler @@ -158,25 +149,26 @@ void main_graph_googlenet(int argc, const char **argv) Graph graph; graph << TargetHint::OPENCL - << Tensor(TensorInfo(TensorShape(224U, 224U, 3U, batches), 1, DataType::F32), DummyAccessor()) + << Tensor(TensorInfo(TensorShape(224U, 224U, 3U, 1U), 1, DataType::F32), + get_input_accessor(image, mean_r, mean_g, mean_b)) << ConvolutionLayer( 7U, 7U, 64U, - get_accessor(data_path, "/cnn_data/googlenet_model/conv1/conv1_7x7_s2_w.npy"), - get_accessor(data_path, "/cnn_data/googlenet_model/conv1/conv1_7x7_s2_b.npy"), + get_weights_accessor(data_path, "/cnn_data/googlenet_model/conv1/conv1_7x7_s2_w.npy"), + get_weights_accessor(data_path, "/cnn_data/googlenet_model/conv1/conv1_7x7_s2_b.npy"), PadStrideInfo(2, 2, 3, 3)) << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)) << PoolingLayer(PoolingLayerInfo(PoolingType::MAX, 3, PadStrideInfo(2, 2, 0, 0, DimensionRoundingType::CEIL))) << NormalizationLayer(NormalizationLayerInfo(NormType::CROSS_MAP, 5, 0.0001f, 0.75f)) << ConvolutionLayer( 1U, 1U, 64U, - get_accessor(data_path, "/cnn_data/googlenet_model/conv2/conv2_3x3_reduce_w.npy"), - get_accessor(data_path, "/cnn_data/googlenet_model/conv2/conv2_3x3_reduce_b.npy"), + get_weights_accessor(data_path, "/cnn_data/googlenet_model/conv2/conv2_3x3_reduce_w.npy"), + get_weights_accessor(data_path, "/cnn_data/googlenet_model/conv2/conv2_3x3_reduce_b.npy"), PadStrideInfo(1, 1, 0, 0)) << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)) << ConvolutionLayer( 3U, 3U, 192U, - get_accessor(data_path, "/cnn_data/googlenet_model/conv2/conv2_3x3_w.npy"), - get_accessor(data_path, "/cnn_data/googlenet_model/conv2/conv2_3x3_b.npy"), + get_weights_accessor(data_path, "/cnn_data/googlenet_model/conv2/conv2_3x3_w.npy"), + get_weights_accessor(data_path, "/cnn_data/googlenet_model/conv2/conv2_3x3_b.npy"), PadStrideInfo(1, 1, 1, 1)) << ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU)) << NormalizationLayer(NormalizationLayerInfo(NormType::CROSS_MAP, 5, 0.0001f, 0.75f)) @@ -195,10 +187,10 @@ void main_graph_googlenet(int argc, const char **argv) << PoolingLayer(PoolingLayerInfo(PoolingType::AVG, 7, PadStrideInfo(1, 1, 0, 0, DimensionRoundingType::CEIL))) << FullyConnectedLayer( 1000U, - get_accessor(data_path, "/cnn_data/googlenet_model/loss3/loss3_classifier_w.npy"), - get_accessor(data_path, "/cnn_data/googlenet_model/loss3/loss3_classifier_b.npy")) + get_weights_accessor(data_path, "/cnn_data/googlenet_model/loss3/loss3_classifier_w.npy"), + get_weights_accessor(data_path, "/cnn_data/googlenet_model/loss3/loss3_classifier_b.npy")) << SoftmaxLayer() - << Tensor(DummyAccessor()); + << Tensor(get_output_accessor(label, 5)); graph.run(); } @@ -206,7 +198,7 @@ void main_graph_googlenet(int argc, const char **argv) /** Main program for Googlenet * * @param[in] argc Number of arguments - * @param[in] argv Arguments ( [optional] Path to the weights folder, [optional] batches ) + * @param[in] argv Arguments ( [optional] Path to the weights folder, [optional] image, [optional] labels ) */ int main(int argc, const char **argv) { diff --git a/utils/GraphUtils.cpp b/utils/GraphUtils.cpp index 15767632c8..9a1ed3d7d9 100644 --- a/utils/GraphUtils.cpp +++ b/utils/GraphUtils.cpp @@ -98,6 +98,9 @@ bool PPMAccessor::access_tensor(ITensor &tensor) // Open PPM file ppm.open(_ppm_path); + ARM_COMPUTE_ERROR_ON_MSG(ppm.width() != tensor.info()->dimension(0) || ppm.height() != tensor.info()->dimension(1), + "Failed to load image file: dimensions [%d,%d] not correct, expected [%d,%d].", ppm.width(), ppm.height(), tensor.info()->dimension(0), tensor.info()->dimension(1)); + // Fill the tensor with the PPM content (BGR) ppm.fill_planar_tensor(tensor, _bgr); diff --git a/utils/GraphUtils.h b/utils/GraphUtils.h index 39b3f115bd..d7d5cd6778 100644 --- a/utils/GraphUtils.h +++ b/utils/GraphUtils.h @@ -175,6 +175,72 @@ public: private: const std::string _filename; }; + +/** Generates appropriate weights accessor according to the specified path + * + * @note If path is empty will generate a DummyAccessor else will generate a NumPyBinLoader + * + * @param[in] path Path to the data files + * @param[in] data_file Relative path to the data files from path + * + * @return An appropriate tensor accessor + */ +inline std::unique_ptr<graph::ITensorAccessor> get_weights_accessor(const std::string &path, const std::string &data_file) +{ + if(path.empty()) + { + return arm_compute::support::cpp14::make_unique<DummyAccessor>(); + } + else + { + return arm_compute::support::cpp14::make_unique<NumPyBinLoader>(path + data_file); + } +} + +/** Generates appropriate input accessor according to the specified ppm_path + * + * @note If ppm_path is empty will generate a DummyAccessor else will generate a PPMAccessor + * + * @param[in] ppm_path Path to PPM file + * @param[in] mean_r Red mean value to be subtracted from red channel + * @param[in] mean_g Green mean value to be subtracted from green channel + * @param[in] mean_b Blue mean value to be subtracted from blue channel + * + * @return An appropriate tensor accessor + */ +inline std::unique_ptr<graph::ITensorAccessor> get_input_accessor(const std::string &ppm_path, float mean_r, float mean_g, float mean_b) +{ + if(ppm_path.empty()) + { + return arm_compute::support::cpp14::make_unique<DummyAccessor>(); + } + else + { + return arm_compute::support::cpp14::make_unique<PPMAccessor>(ppm_path, true, mean_r, mean_g, mean_b); + } +} + +/** Generates appropriate output accessor according to the specified labels_path + * + * @note If labels_path is empty will generate a DummyAccessor else will generate a TopNPredictionsAccessor + * + * @param[in] labels_path Path to labels text file + * @param[in] top_n (Optional) Number of output classes to print + * @param[out] output_stream (Optional) Output stream + * + * @return An appropriate tensor accessor + */ +inline std::unique_ptr<graph::ITensorAccessor> get_output_accessor(const std::string &labels_path, size_t top_n = 5, std::ostream &output_stream = std::cout) +{ + if(labels_path.empty()) + { + return arm_compute::support::cpp14::make_unique<DummyAccessor>(); + } + else + { + return arm_compute::support::cpp14::make_unique<TopNPredictionsAccessor>(labels_path, top_n, output_stream); + } +} } // namespace graph_utils } // namespace arm_compute diff --git a/utils/Utils.h b/utils/Utils.h index 1f3d971917..28382f47e4 100644 --- a/utils/Utils.h +++ b/utils/Utils.h @@ -410,6 +410,20 @@ public: } } + /** Return the width of the currently open PPM file. + */ + unsigned int width() const + { + return _width; + } + + /** Return the height of the currently open PPM file. + */ + unsigned int height() const + { + return _height; + } + private: std::ifstream _fs; unsigned int _width, _height; |