Class FPN<T>

Namespace

AiDotNet.ComputerVision.Detection.Necks

Assembly

AiDotNet.dll

Feature Pyramid Network (FPN) for multi-scale feature fusion.

public class FPN<T> : NeckBase<T>

Type Parameters

T

The numeric type used for calculations.

Inheritance

object

NeckBase<T>

FPN<T>

Inherited Members

NeckBase<T>.NumOps

NeckBase<T>.IsTrainingMode

NeckBase<T>.SetTrainingMode(bool)

NeckBase<T>.ValidateFeatures(List<Tensor<T>>, int[])

NeckBase<T>.Upsample2x(Tensor<T>)

NeckBase<T>.Downsample2x(Tensor<T>)

NeckBase<T>.Conv1x1(Tensor<T>, Tensor<T>, Tensor<T>)

NeckBase<T>.Add(Tensor<T>, Tensor<T>)

object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.MemberwiseClone()

object.ReferenceEquals(object, object)

object.ToString()

Remarks

For Beginners: FPN creates a feature pyramid by combining features from different backbone levels using a top-down pathway. High-resolution features from earlier layers are enriched with semantic information from deeper layers.

Key features: - Top-down pathway with lateral connections - 1x1 convolutions to match channel dimensions - Simple element-wise addition for fusion - Fast and memory efficient

Reference: Lin et al., "Feature Pyramid Networks for Object Detection", CVPR 2017

Constructors

FPN(NeckConfig)

Creates FPN from a configuration object.

public FPN(NeckConfig config)

Parameters

config NeckConfig

Neck configuration.

FPN(int[], int)

Creates a new Feature Pyramid Network.

public FPN(int[] inputChannels, int outputChannels = 256)

Parameters

inputChannels int[]

Channel counts from backbone at each level.

outputChannels int

Output channel count for all levels (default 256).

Properties

Name

Name of this neck architecture.

public override string Name { get; }

Property Value

string

NumLevels

Number of feature levels output by the neck.

public override int NumLevels { get; }

Property Value

int

OutputChannels

Number of output channels for all feature levels.

public override int OutputChannels { get; }

Property Value

int

Remarks

Necks typically project all feature levels to the same number of channels (e.g., 256) to simplify the detection head.

Methods

Forward(List<Tensor<T>>)

Performs multi-scale feature fusion.

public override List<Tensor<T>> Forward(List<Tensor<T>> features)

Parameters

features List<Tensor<T>>

List of feature maps from the backbone, ordered from highest to lowest resolution.

Returns

List<Tensor<T>>

Fused feature maps at multiple scales.

Remarks

For Beginners: This method takes the raw features from the backbone and combines them across scales. After fusion, each feature level "knows about" features from other scales, making detection more accurate.

GetParameterCount()

Gets the total number of parameters in the neck.

public override long GetParameterCount()

Returns

long

Number of trainable parameters.

ReadParameters(BinaryReader)

Reads parameters from a binary reader for deserialization.

public override void ReadParameters(BinaryReader reader)

Parameters

reader BinaryReader

The binary reader to read from.

WriteParameters(BinaryWriter)

Writes all parameters to a binary writer for serialization.

public override void WriteParameters(BinaryWriter writer)

Parameters

writer BinaryWriter

The binary writer to write to.