develop/amd_openvx/openvx/include/VX/vx_khr_nn.h Source File#
vx_khr_nn.h
Go to the documentation of this file.
58 VX_KERNEL_FULLY_CONNECTED_LAYER = VX_KERNEL_BASE(VX_ID_KHRONOS, VX_LIBRARY_KHR_NN_EXTENSION) + 0x1,
78 VX_KERNEL_DECONVOLUTION_LAYER = VX_KERNEL_BASE(VX_ID_KHRONOS, VX_LIBRARY_KHR_NN_EXTENSION) + 0x7,
82 VX_KERNEL_LOCAL_RESPONSE_NORMALIZATION_LAYER = VX_KERNEL_BASE(VX_ID_KHRONOS, VX_LIBRARY_KHR_NN_EXTENSION) + 0x8,
108 };
121 };
130 VX_NN_NORMALIZATION_SAME_MAP = VX_ENUM_BASE(VX_ID_KHRONOS, VX_ENUM_NN_NORMALIZATION_TYPE) + 0x0,
132 VX_NN_NORMALIZATION_ACROSS_MAPS = VX_ENUM_BASE(VX_ID_KHRONOS, VX_ENUM_NN_NORMALIZATION_TYPE) + 0x1,
155 VX_NN_ACTIVATION_LOGISTIC = VX_ENUM_BASE(VX_ID_KHRONOS, VX_ENUM_NN_ACTIVATION_FUNCTION_TYPE) + 0x0,
156 VX_NN_ACTIVATION_HYPERBOLIC_TAN = VX_ENUM_BASE(VX_ID_KHRONOS, VX_ENUM_NN_ACTIVATION_FUNCTION_TYPE) + 0x1,
158 VX_NN_ACTIVATION_BRELU = VX_ENUM_BASE(VX_ID_KHRONOS, VX_ENUM_NN_ACTIVATION_FUNCTION_TYPE) + 0x3,
159 VX_NN_ACTIVATION_SOFTRELU = VX_ENUM_BASE(VX_ID_KHRONOS, VX_ENUM_NN_ACTIVATION_FUNCTION_TYPE) + 0x4,
161 VX_NN_ACTIVATION_SQUARE = VX_ENUM_BASE(VX_ID_KHRONOS, VX_ENUM_NN_ACTIVATION_FUNCTION_TYPE) + 0x6,
163 VX_NN_ACTIVATION_LINEAR = VX_ENUM_BASE(VX_ID_KHRONOS, VX_ENUM_NN_ACTIVATION_FUNCTION_TYPE) + 0x8,
252 VX_API_ENTRY vx_node VX_API_CALL vxConvolutionLayer(vx_graph graph, vx_tensor inputs, vx_tensor weights, vx_tensor biases, const vx_nn_convolution_params_t *convolution_params, vx_size size_of_convolution_params, vx_tensor outputs);
280 VX_API_ENTRY vx_node VX_API_CALL vxFullyConnectedLayer(vx_graph graph, vx_tensor inputs, vx_tensor weights, vx_tensor biases, vx_enum overflow_policy, vx_enum rounding_policy, vx_tensor outputs);
307 VX_API_ENTRY vx_node VX_API_CALL vxPoolingLayer(vx_graph graph, vx_tensor inputs, vx_enum pooling_type,
334 VX_API_ENTRY vx_node VX_API_CALL vxSoftmaxLayer(vx_graph graph, vx_tensor inputs, vx_tensor outputs);
355 VX_API_ENTRY vx_node VX_API_CALL vxLocalResponseNormalizationLayer(vx_graph graph, vx_tensor inputs, vx_enum type,
378 VX_API_ENTRY vx_node VX_API_CALL vxActivationLayer(vx_graph graph, vx_tensor inputs, vx_enum function, vx_float32 a,vx_float32 b, vx_tensor outputs);
399 VX_API_ENTRY vx_node VX_API_CALL vxROIPoolingLayer(vx_graph graph, vx_tensor input_data, vx_tensor input_rois, const vx_nn_roi_pool_params_t *roi_pool_params, vx_size size_of_roi_params, vx_tensor output_arr);
446 VX_API_ENTRY vx_node VX_API_CALL vxDeconvolutionLayer(vx_graph graph, vx_tensor inputs, vx_tensor weights, vx_tensor biases, const vx_nn_deconvolution_params_t *deconvolution_params, vx_size size_of_deconv_params, vx_tensor outputs);
VX_API_ENTRY vx_node VX_API_CALL vxDeconvolutionLayer(vx_graph graph, vx_tensor inputs, vx_tensor weights, vx_tensor biases, const vx_nn_deconvolution_params_t *deconvolution_params, vx_size size_of_deconv_params, vx_tensor outputs)
[Graph] Creates a Convolutional Network Deconvolution Layer Node.
VX_API_ENTRY vx_node VX_API_CALL vxConvolutionLayer(vx_graph graph, vx_tensor inputs, vx_tensor weights, vx_tensor biases, const vx_nn_convolution_params_t *convolution_params, vx_size size_of_convolution_params, vx_tensor outputs)
[Graph] Creates a Convolutional Network Convolution Layer Node.
#define VX_LIBRARY_KHR_NN_EXTENSION
The Neural Network Extension Library Set.
Definition: vx_khr_nn.h:45
struct _vx_nn_deconvolution_params_t vx_nn_deconvolution_params_t
Input parameters for a deconvolution operation.
vx_nn_type_e
The type enumeration lists all NN extension types.
Definition: vx_khr_nn.h:170
struct _vx_nn_convolution_params_t vx_nn_convolution_params_t
Input parameters for a convolution operation.
struct _vx_nn_roi_pool_params_t vx_nn_roi_pool_params_t
Input parameters for ROI pooling operation.
VX_API_ENTRY vx_node VX_API_CALL vxROIPoolingLayer(vx_graph graph, vx_tensor input_data, vx_tensor input_rois, const vx_nn_roi_pool_params_t *roi_pool_params, vx_size size_of_roi_params, vx_tensor output_arr)
[Graph] Creates a Convolutional Network ROI pooling node
vx_nn_pooling_type_e
The Neural Network pooling type list.
Definition: vx_khr_nn.h:116
VX_API_ENTRY vx_node VX_API_CALL vxLocalResponseNormalizationLayer(vx_graph graph, vx_tensor inputs, vx_enum type, vx_size normalization_size, vx_float32 alpha, vx_float32 beta, vx_float32 bias, vx_tensor outputs)
[Graph] Creates a Convolutional Network Local Response Normalization Layer Node. This function is opt...
vx_nn_norm_type_e
The Neural Network normalization type list.
Definition: vx_khr_nn.h:128
VX_API_ENTRY vx_node VX_API_CALL vxSoftmaxLayer(vx_graph graph, vx_tensor inputs, vx_tensor outputs)
[Graph] Creates a Convolutional Network Softmax Layer Node.
VX_API_ENTRY vx_node VX_API_CALL vxActivationLayer(vx_graph graph, vx_tensor inputs, vx_enum function, vx_float32 a, vx_float32 b, vx_tensor outputs)
[Graph] Creates a Convolutional Network Activation Layer Node. The function operate a specific functi...
VX_API_ENTRY vx_node VX_API_CALL vxFullyConnectedLayer(vx_graph graph, vx_tensor inputs, vx_tensor weights, vx_tensor biases, vx_enum overflow_policy, vx_enum rounding_policy, vx_tensor outputs)
[Graph] Creates a Fully connected Convolutional Network Layer Node.
vx_kernel_nn_ext_e
The list of Neural Network Extension Kernels.
Definition: vx_khr_nn.h:50
VX_API_ENTRY vx_node VX_API_CALL vxPoolingLayer(vx_graph graph, vx_tensor inputs, vx_enum pooling_type, vx_size pooling_size_x, vx_size pooling_size_y, vx_size pooling_padding_x, vx_size pooling_padding_y, vx_enum rounding, vx_tensor outputs)
[Graph] Creates a Convolutional Network Pooling Layer Node.
vx_nn_activation_function_e
The Neural Network activation functions list.
Definition: vx_khr_nn.h:154
@ VX_TYPE_NN_CONVOLUTION_PARAMS
A vx_nn_convolution_params_t.
Definition: vx_khr_nn.h:171
@ VX_TYPE_NN_DECONVOLUTION_PARAMS
A vx_nn_deconvolution_params_t.
Definition: vx_khr_nn.h:172
@ VX_TYPE_NN_ROI_POOL_PARAMS
A vx_nn_roi_pool_params_t.
Definition: vx_khr_nn.h:173
@ VX_NN_NORMALIZATION_ACROSS_MAPS
Normalization is done across different IFMs.
Definition: vx_khr_nn.h:132
@ VX_NN_NORMALIZATION_SAME_MAP
normalization is done on same IFM
Definition: vx_khr_nn.h:130
@ VX_NN_DS_SIZE_ROUNDING_CEILING
ceil rounding
Definition: vx_khr_nn.h:107
@ VX_KERNEL_LOCAL_RESPONSE_NORMALIZATION_LAYER
The Neural Network Extension local response normalization Kernel (with bias).
Definition: vx_khr_nn.h:82
@ VX_KERNEL_ROI_POOLING_LAYER
The Neural Network POI Pooling Kernel.
Definition: vx_khr_nn.h:74
@ VX_KERNEL_DECONVOLUTION_LAYER
The Neural Network Extension Deconvolution Kernel.
Definition: vx_khr_nn.h:78
@ VX_KERNEL_CONVOLUTION_LAYER
The Neural Network Extension convolution Kernel.
Definition: vx_khr_nn.h:54
@ VX_KERNEL_SOFTMAX_LAYER
The Neural Network Extension softmax Kernel.
Definition: vx_khr_nn.h:66
@ VX_KERNEL_FULLY_CONNECTED_LAYER
The Neural Network Extension fully connected Kernel.
Definition: vx_khr_nn.h:58
@ VX_KERNEL_POOLING_LAYER
The Neural Network Extension pooling Kernel.
Definition: vx_khr_nn.h:62
@ VX_KERNEL_ACTIVATION_LAYER
The Neural Network Extension activation Kernel.
Definition: vx_khr_nn.h:70
vx_enum overflow_policy
A VX_TYPE_ENUM of the vx_convert_policy_e enumeration.
Definition: vx_khr_nn.h:183
vx_size padding_y
Number of elements added at each side in the y dimension of the input.
Definition: vx_khr_nn.h:182
vx_size dilation_y
“inflate” the kernel by inserting zeros between the kernel elements in the y direction....
Definition: vx_khr_nn.h:187
vx_enum rounding_policy
A VX_TYPE_ENUM of the vx_round_policy_e enumeration.
Definition: vx_khr_nn.h:184
vx_size padding_x
Number of elements added at each side in the x dimension of the input.
Definition: vx_khr_nn.h:181
vx_size dilation_x
“inflate” the kernel by inserting zeros between the kernel elements in the x direction....
Definition: vx_khr_nn.h:186
vx_enum down_scale_size_rounding
Rounding method for calculating output dimensions. See vx_nn_rounding_type_e
Definition: vx_khr_nn.h:185
Input parameters for a deconvolution operation.
Definition: vx_khr_nn.h:195
vx_enum rounding_policy
A VX_TYPE_ENUM of the vx_round_policy_e enumeration.
Definition: vx_khr_nn.h:199
vx_size padding_y
Number of elements subtracted at each side in the y dimension of the output.
Definition: vx_khr_nn.h:197
vx_size a_y
user-specified quantity used to distinguish between the different possible output sizes.
Definition: vx_khr_nn.h:201
vx_size padding_x
Number of elements subtracted at each side in the x dimension of the output.
Definition: vx_khr_nn.h:196
vx_enum overflow_policy
A VX_TYPE_ENUM of the vx_convert_policy_e enumeration.
Definition: vx_khr_nn.h:198
vx_size a_x
user-specified quantity used to distinguish between the different possible output sizes.
Definition: vx_khr_nn.h:200
vx_enum pool_type
Of type vx_nn_pooling_type_e. Only VX_NN_POOLING_MAX pooling is supported.
Definition: vx_khr_nn.h:209
The top level OpenVX Header.
size_t vx_size
A wrapper of size_t to keep the naming convention uniform.
Definition: vx_types.h:157
struct _vx_tensor_t * vx_tensor
The multidimensional data object (Tensor).
Definition: vx_types.h:287
int32_t vx_enum
Sets the standard enumeration type size to be a fixed quantity.
Definition: vx_types.h:152
#define VX_ENUM_BASE(vendor, id)
Defines the manner in which to combine the Vendor and Object IDs to get the base value of the enumera...
Definition: vx_types.h:550
#define VX_KERNEL_BASE(vendor, lib)
Defines the manner in which to combine the Vendor and Library IDs to get the base value of the enumer...
Definition: vx_types.h:540