Partial List of Implemented Layers

Detailed Description

This subsection of dnn module contains information about built-in layers and their descriptions.

Classes listed here, in fact, provides C++ API for creating instances of built-in layers. In addition to this way of layers instantiation, there is a more common factory API (see Utilities for New Layers Registration), it allows to create layers dynamically (by name) and register new ones. You can use both API, but factory API is less convenient for native C++ programming and basically designed for use inside importers (see readNetFromTensorflow()).

Built-in layers reproduce the functionality of the corresponding ONNX and TensorFlow operators. The following layers are among the core building blocks used to assemble imported networks:

• Convolution
• Deconvolution
• Pooling
• InnerProduct
• TanH, ReLU, Sigmoid, BNLL, Power, AbsVal
• Softmax
• Reshape, Flatten, Slice, Split
• LRN
• MVN
• Dropout (since it does nothing on forward pass -))

Classes
class	cv::dnn::AbsLayer
class	cv::dnn::AccumLayer
class	cv::dnn::AcoshLayer
class	cv::dnn::AcosLayer
struct	cv::dnn::ActivationInt8Params
class	cv::dnn::ActivationLayer
class	cv::dnn::ActivationLayerInt8
class	cv::dnn::AffineGridLayer
class	cv::dnn::ArgLayer
	ArgMax/ArgMin layer. More...
class	cv::dnn::AsinhLayer
class	cv::dnn::AsinLayer
class	cv::dnn::AtanhLayer
class	cv::dnn::AtanLayer
class	cv::dnn::AttentionLayer
class	cv::dnn::AttentionOnnxAiLayer
class	cv::dnn::AveragePoolLayer
class	cv::dnn::BaseConvolutionLayer
class	cv::dnn::BatchNorm2Layer
class	cv::dnn::BatchNormLayer
class	cv::dnn::BatchNormLayerInt8
class	cv::dnn::BitShiftLayer
class	cv::dnn::BlackmanWindowLayer
class	cv::dnn::BlankLayer
class	cv::dnn::BNLLLayer
class	cv::dnn::Cast2Layer
class	cv::dnn::CastLayer
class	cv::dnn::CeilLayer
class	cv::dnn::CeluLayer
class	cv::dnn::CenterCropPadLayer
class	cv::dnn::ChannelsPReLULayer
class	cv::dnn::ClipLayer
class	cv::dnn::CompareLayer
class	cv::dnn::Concat2Layer
class	cv::dnn::ConcatLayer
class	cv::dnn::ConstantOfShapeLayer
class	cv::dnn::ConstLayer
class	cv::dnn::Conv2Int8Layer
struct	cv::dnn::Conv2Int8Params
class	cv::dnn::Conv2Layer
class	cv::dnn::ConvolutionLayer
class	cv::dnn::ConvolutionLayerInt8
class	cv::dnn::ConvTranspose2Layer
class	cv::dnn::CorrelationLayer
class	cv::dnn::CoshLayer
class	cv::dnn::CosLayer
class	cv::dnn::CropAndResizeLayer
class	cv::dnn::CropLayer
class	cv::dnn::CumSumLayer
class	cv::dnn::DataAugmentationLayer
class	cv::dnn::DeconvolutionLayer
class	cv::dnn::DepthToSpaceLayer
class	cv::dnn::DequantizeLayer
class	cv::dnn::DequantizeLinearLayer
class	cv::dnn::DetectionOutputLayer
	Detection output layer. More...
class	cv::dnn::DetLayer
class	cv::dnn::DFTLayer
class	cv::dnn::EinsumLayer
	This function performs array summation based on the Einstein summation convention. The function allows for concise expressions of various mathematical operations using subscripts. More...
class	cv::dnn::Eltwise2Int8Layer
struct	cv::dnn::Eltwise2Int8Params
class	cv::dnn::EltwiseLayer
	Element wise operation on inputs. More...
class	cv::dnn::EltwiseLayerInt8
class	cv::dnn::ELULayer
class	cv::dnn::ErfLayer
class	cv::dnn::Expand2Layer
class	cv::dnn::ExpandLayer
class	cv::dnn::ExpLayer
class	cv::dnn::EyeLikeLayer
class	cv::dnn::FlattenLayer
class	cv::dnn::FloorLayer
class	cv::dnn::FlowWarpLayer
class	cv::dnn::Gather2Layer
class	cv::dnn::GatherElementsLayer
	GatherElements layer GatherElements takes two inputs data and indices of the same rank r >= 1 and an optional attribute axis and works such that: output[i][j][k] = data[index[i][j][k]][j][k] if axis = 0 and r = 3 output[i][j][k] = data[i][index[i][j][k]][k] if axis = 1 and r = 3 output[i][j][k] = data[i][j][index[i][j][k]] if axis = 2 and r = 3. More...
class	cv::dnn::GatherLayer
	Gather layer. More...
class	cv::dnn::GatherNDLayer
	GatherND layer. More...
class	cv::dnn::GeluApproximationLayer
class	cv::dnn::GeluLayer
class	cv::dnn::GemmLayer
class	cv::dnn::GlobalAveragePoolLayer
class	cv::dnn::GridSampleLayer
class	cv::dnn::GroupNormLayer
class	cv::dnn::GRULayer
	GRU recurrent one-layer. More...
class	cv::dnn::HammingWindowLayer
class	cv::dnn::HannWindowLayer
class	cv::dnn::HardmaxLayer
class	cv::dnn::HardSigmoidLayer
class	cv::dnn::HardSwishLayer
class	cv::dnn::IfLayer
struct	cv::dnn::InnerProductInt8Params
class	cv::dnn::InnerProductLayer
class	cv::dnn::InnerProductLayerInt8
class	cv::dnn::InstanceNormLayer
class	cv::dnn::InterpLayer
	Bilinear resize layer from https://github.com/cdmh/deeplab-public-ver2. More...
class	cv::dnn::IsInfLayer
class	cv::dnn::IsNaNLayer
class	cv::dnn::LayerNorm2Layer
class	cv::dnn::LayerNormLayer
class	cv::dnn::LogLayer
class	cv::dnn::LoopLayer
class	cv::dnn::LRNLayer
class	cv::dnn::LSTM2Layer
class	cv::dnn::LSTMLayer
	LSTM recurrent layer. More...
class	cv::dnn::MatMulInt8Layer
struct	cv::dnn::MatMulInt8Params
class	cv::dnn::MatMulLayer
class	cv::dnn::MaxPoolLayer
class	cv::dnn::MaxUnpoolLayer
class	cv::dnn::MishLayer
class	cv::dnn::MVNLayer
class	cv::dnn::NaryEltwiseLayer
class	cv::dnn::NegativeLogLikelihoodLossLayer
class	cv::dnn::NonMaxSuppressionLayer
class	cv::dnn::NonZeroLayer
class	cv::dnn::NormalizeBBoxLayer
	\( L_p \) - normalization layer. More...
class	cv::dnn::NotLayer
class	cv::dnn::OneHotLayer
class	cv::dnn::Pad2Layer
class	cv::dnn::PaddingLayer
	Adds extra values for specific axes. More...
class	cv::dnn::PermuteLayer
class	cv::dnn::Pool2Int8Layer
struct	cv::dnn::Pool2Int8Params
class	cv::dnn::PoolingLayer
class	cv::dnn::PoolingLayerInt8
class	cv::dnn::PowerLayer
class	cv::dnn::PriorBoxLayer
class	cv::dnn::ProposalLayer
class	cv::dnn::QuantizeLayer
class	cv::dnn::QuantizeLinearLayer
class	cv::dnn::RandomNormalLikeLayer
class	cv::dnn::RangeLayer
class	cv::dnn::ReciprocalLayer
class	cv::dnn::Reduce2Layer
class	cv::dnn::ReduceLayer
class	cv::dnn::RegionLayer
class	cv::dnn::ReLU6Layer
class	cv::dnn::ReLULayer
class	cv::dnn::ReorgLayer
class	cv::dnn::RequantizeLayer
class	cv::dnn::Reshape2Layer
class	cv::dnn::ReshapeLayer
class	cv::dnn::Resize2Layer
class	cv::dnn::ResizeLayer
	Resize input 4-dimensional blob by nearest neighbor or bilinear strategy. More...
class	cv::dnn::RMSNormLayer
class	cv::dnn::RNNLayer
	Classical recurrent layer. More...
class	cv::dnn::RoiAlignLayer
class	cv::dnn::RotaryEmbeddingLayer
class	cv::dnn::RoundLayer
class	cv::dnn::ScaleLayer
class	cv::dnn::ScaleLayerInt8
class	cv::dnn::ScatterLayer
class	cv::dnn::ScatterNDLayer
class	cv::dnn::SDPALayer
class	cv::dnn::SeluLayer
class	cv::dnn::ShapeLayer
class	cv::dnn::ShiftLayer
class	cv::dnn::ShiftLayerInt8
class	cv::dnn::ShrinkLayer
class	cv::dnn::ShuffleChannelLayer
class	cv::dnn::SigmoidLayer
class	cv::dnn::SignLayer
class	cv::dnn::SinhLayer
class	cv::dnn::SinLayer
class	cv::dnn::SizeLayer
class	cv::dnn::Slice2Layer
class	cv::dnn::SliceLayer
class	cv::dnn::SoftmaxCrossEntropyLossLayer
class	cv::dnn::SoftmaxLayer
class	cv::dnn::SoftmaxLayerInt8
class	cv::dnn::SoftplusLayer
class	cv::dnn::SoftsignLayer
class	cv::dnn::SpaceToDepthLayer
class	cv::dnn::Split2Layer
class	cv::dnn::SplitLayer
class	cv::dnn::SqrtLayer
class	cv::dnn::SqueezeLayer
class	cv::dnn::SwishLayer
class	cv::dnn::TanHLayer
class	cv::dnn::TanLayer
class	cv::dnn::ThresholdedReluLayer
class	cv::dnn::Tile2Layer
class	cv::dnn::TileLayer
class	cv::dnn::TopK2Layer
class	cv::dnn::TopKLayer
class	cv::dnn::TransformLayoutLayer
class	cv::dnn::TransposeLayer
class	cv::dnn::TriluLayer
class	cv::dnn::UniqueLayer
class	cv::dnn::UnsqueezeLayer

Typedefs
typedef void(*	cv::dnn::ActivationFunc) (const void *input, void *output, size_t len, const float *params)

Enumerations
enum	cv::dnn::ActivationType {   cv::dnn::ACTIV_NONE = 0 ,   cv::dnn::ACTIV_MISH ,   cv::dnn::ACTIV_SWISH ,   cv::dnn::ACTIV_SIGMOID ,   cv::dnn::ACTIV_TANH ,   cv::dnn::ACTIV_ELU ,   cv::dnn::ACTIV_HARDSWISH ,   cv::dnn::ACTIV_HARDSIGMOID ,   cv::dnn::ACTIV_GELU ,   cv::dnn::ACTIV_GELU_APPROX ,   cv::dnn::ACTIV_RELU ,   cv::dnn::ACTIV_CLIP }
enum	cv::dnn::AutoPadding {   cv::dnn::AUTO_PAD_NONE = 0 ,   cv::dnn::AUTO_PAD_SAME_UPPER = 1 ,   cv::dnn::AUTO_PAD_SAME_LOWER = 2 ,   cv::dnn::AUTO_PAD_VALID = 3 }
enum	cv::dnn::LossReduction {   cv::dnn::LOSS_REDUCTION_NONE = 0 ,   cv::dnn::LOSS_REDUCTION_MEAN = 1 ,   cv::dnn::LOSS_REDUCTION_SUM = 2 }

Functions
ActivationFunc	cv::dnn::getActivationFunc (int activationType)

Typedef Documentation

ActivationFunc

typedef void(* cv::dnn::ActivationFunc) (const void *input, void *output, size_t len, const float *params)

#include <opencv2/dnn/all_layers.hpp>

Enumeration Type Documentation

ActivationType

enum cv::dnn::ActivationType

#include <opencv2/dnn/all_layers.hpp>

Activation type enumeration for dispatched activation function retrieval.

Enumerator
ACTIV_NONE  Python: cv.dnn.ACTIV_NONE
ACTIV_MISH  Python: cv.dnn.ACTIV_MISH
ACTIV_SWISH  Python: cv.dnn.ACTIV_SWISH
ACTIV_SIGMOID  Python: cv.dnn.ACTIV_SIGMOID
ACTIV_TANH  Python: cv.dnn.ACTIV_TANH
ACTIV_ELU  Python: cv.dnn.ACTIV_ELU
ACTIV_HARDSWISH  Python: cv.dnn.ACTIV_HARDSWISH
ACTIV_HARDSIGMOID  Python: cv.dnn.ACTIV_HARDSIGMOID
ACTIV_GELU  Python: cv.dnn.ACTIV_GELU
ACTIV_GELU_APPROX  Python: cv.dnn.ACTIV_GELU_APPROX
ACTIV_RELU  Python: cv.dnn.ACTIV_RELU
ACTIV_CLIP  Python: cv.dnn.ACTIV_CLIP

AutoPadding

enum cv::dnn::AutoPadding

#include <opencv2/dnn/all_layers.hpp>

Enumerator
AUTO_PAD_NONE  Python: cv.dnn.AUTO_PAD_NONE
AUTO_PAD_SAME_UPPER  Python: cv.dnn.AUTO_PAD_SAME_UPPER
AUTO_PAD_SAME_LOWER  Python: cv.dnn.AUTO_PAD_SAME_LOWER
AUTO_PAD_VALID  Python: cv.dnn.AUTO_PAD_VALID

LossReduction

enum cv::dnn::LossReduction

#include <opencv2/dnn/all_layers.hpp>

Enumerator
LOSS_REDUCTION_NONE  Python: cv.dnn.LOSS_REDUCTION_NONE
LOSS_REDUCTION_MEAN  Python: cv.dnn.LOSS_REDUCTION_MEAN
LOSS_REDUCTION_SUM  Python: cv.dnn.LOSS_REDUCTION_SUM

Function Documentation

getActivationFunc()

ActivationFunc cv::dnn::getActivationFunc

(

int

activationType

)

#include <opencv2/dnn/all_layers.hpp>

Returns a platform-optimized activation function pointer for the given type. The returned function is selected via CPU dispatch for the best available ISA.