forward

Run forward pass on Mask R-CNN network

Since R2021b

Syntax

[dlRPNScores,dlRPNReg,dlProposals,dlBoxScores,dlBoxReg,dlMasks] = forward(detector,dlX)

[dlRPNScores,dlRPNReg,dlProposals,dlBoxScores,dlBoxReg,dlMasks,state] = forward(detector,dlX)

Description

[dlRPNScores,dlRPNReg,dlProposals,dlBoxScores,dlBoxReg,dlMasks] = forward(detector,dlX) calculates features of the image dlX from the output layers of the Mask R-CNN object detector.

[dlRPNScores,dlRPNReg,dlProposals,dlBoxScores,dlBoxReg,dlMasks,state] = forward(detector,dlX) also returns the state information of the network. Use the state to update the network parameters.

Note

This function requires the Computer Vision Toolbox™ Model for Mask R-CNN Instance Segmentation. You can install the Computer Vision Toolbox Model for Mask R-CNN Instance Segmentation from Add-On Explorer. For more information about installing add-ons, see Get and Manage Add-Ons. To run this function, you will require the Deep Learning Toolbox™.

Examples

Calculate Features of Color Image

Load a pretrained Mask R-CNN object detector.

detector = maskrcnn("resnet50-coco");

Read an image to use for training, and convert the image to a formatted dlarray object.

I = imread("visionteam.jpg");
dlX = dlarray(single(I),"SSCB"); 

Calculate features of the training image.

[dlRPNScores,dlRPNReg,dlProposals,dlBoxScores,dlBoxReg,dlMasks] = forward(detector,dlX);

Input Arguments

detector — Mask R-CNN object detector

maskrcnn object

Mask R-CNN object detector, specified as a maskrcnn object.

dlX — Training data

formatted dlarray object

Training data, specified as a formatted dlarray (Deep Learning Toolbox) object containing real, nonsparse data. The dimension labels of the data must be "SSCB".

Output Arguments

dlRPNScores — RPN classification output

dlarray object

RPN classification output after the softmax operation, returned as a dlarray object of size h-by-w-by-numAnchors-by-B. The feature map has spatial size h-by-w. B is the number of images in the batch.

dlRPNReg — RPN regression output

dlarray object

RPN regression output, returned as a dlarray object of size h-by-w-by-(4⁢×numAnchors)-by-B. The feature map has spatial size h-by-w. B is the number of images in the batch.

dlProposals — Region proposals

dlarray object

Region proposals, returned as a dlarray object of size 5-by-numProposals-by-1-by-1. Each column of the proposals contains a box proposal in the format [xStart, yStart, xEnd, yEnd, batchIdx].

dlBoxScores — Detection network classification output

dlarray object

Detection network classification output after the softmax operation, returned as a dlarray object of size 1-by-1-by-(numClasses+1)-by-numProposals.

dlBoxReg — Detection network regression output

dlarray object

Detection network regression output, returned as a dlarray object of size 1-by-1-by-(4×numClasses)-by-numProposals.

dlMasks — Mask segmentation output

dlarray object

Mask segmentation output, returned as a dlarray object of size h_mask-by-w_mask-by-numClasses-by-numProposals. The mask segmentation output has spatial size h_mask-by-w_mask.

state — Updated network state

table

Updated network state, returned as a table. The network state is a table with three columns:

• Layer – Layer name, returned as a string scalar.

• Parameter – Parameter name, returned as a string scalar.

• Value – Value of parameter, returned as a numeric array object.

The network state contains information remembered by the network between iterations.

Version History

Introduced in R2021b

R2022a: Features returned in different format

The forward function now returns the feature maps for the six output layers in separate output arguments. Before, the function returned the feature maps as a 1-by-6 cell array in a single output argument.