Table of Contents

Class MaskHead<T>

Mask prediction head for instance segmentation.

public class MaskHead<T>

Type Parameters

T

The numeric type used for calculations.

Inheritance
MaskHead<T>
Inherited Members

Remarks

For Beginners: The mask head takes RoI-pooled features and predicts a binary segmentation mask for each class. It typically uses a series of convolutional layers followed by a transposed convolution for upsampling.

Key features: - Multiple convolutional layers for feature processing - Upsampling via transposed convolution - Per-class mask prediction - Configurable mask resolution

Constructors

MaskHead(int, int, int)

Creates a new mask head.

public MaskHead(int inChannels, int numClasses, int maskResolution = 28)

Parameters

inChannels int

Number of input channels from RoI features.

numClasses int

Number of classes to predict.

maskResolution int

Output mask resolution.

Methods

Forward(Tensor<T>)

Forward pass to predict masks from RoI features.

public Tensor<T> Forward(Tensor<T> roiFeatures)

Parameters

roiFeatures Tensor<T>

RoI-pooled features [num_rois, channels, height, width].

Returns

Tensor<T>

Mask predictions [num_rois, num_classes, mask_h, mask_w].

GetParameterCount()

Gets the total parameter count.

public long GetParameterCount()

Returns

long

PredictMask(Tensor<T>, int)

Predicts mask for a single RoI and class.

public Tensor<T> PredictMask(Tensor<T> roiFeatures, int classId)

Parameters

roiFeatures Tensor<T>

Features for single RoI [1, channels, h, w].

classId int

Class ID to predict mask for.

Returns

Tensor<T>

Binary mask [mask_h, mask_w].

ReadParameters(BinaryReader)

Reads parameters from binary reader.

public void ReadParameters(BinaryReader reader)

Parameters

reader BinaryReader

WriteParameters(BinaryWriter)

Writes parameters to binary writer.

public void WriteParameters(BinaryWriter writer)

Parameters

writer BinaryWriter