regionProposalLayer

(Not recommended) Region proposal layer for Faster R-CNN

RegionProposalLayer is not recommended. Instead, use a different type of object detector, such as a yoloxObjectDetector or yolov4ObjectDetector detector. For more information, see Version History.

Description

A region proposal layer outputs bounding boxes around potential objects in an image as part of the region proposal network (RPN) within Faster R-CNN. These outputs are further refined by additional layers within Faster R-CNN to produce the final object detection results.

There are two inputs to this layer:

• 'scores' — The classification scores produced by the RPN classification branch

• 'boxDeltas' — The bounding box deltas produced by the RPN regression branch

Use the input names when connecting or disconnecting the region proposal layer to other layers using connectLayers (Deep Learning Toolbox) or disconnectLayers (Deep Learning Toolbox) (requires Deep Learning Toolbox™).

Creation

Syntax

layer = regionProposalLayer(anchorBoxes)

layer = regionProposalLayer(anchorBoxes,'Name',Name)

Description

layer = regionProposalLayer(anchorBoxes) creates a region proposal layer for building Faster R-CNN object detection networks, and sets the AnchorBoxes property.

layer = regionProposalLayer(anchorBoxes,'Name',Name) creates a region proposal layer and sets the optional Name property.

Properties

AnchorBoxes — Anchor boxes

M-by-2 matrix

Anchor boxes, specified as an M-by-2 matrix defining the [height width] of M anchor boxes.

Anchor boxes are predefined bounding box templates of fixed size. The size of each anchor box is typically determined based on a priori knowledge of the scale and aspect ratio of objects in the training dataset. An RPN network is trained to predict the translation and rescaling needed to align the anchor boxes with the ground truth bounding boxes. [1]

Name — Layer name

"" (default) | character vector | string scalar

Layer name, specified as a character vector or string scalar. For Layer array input, the trainnet (Deep Learning Toolbox) and dlnetwork (Deep Learning Toolbox) functions automatically assign names to layers with the name "".

The RegionProposalLayer object stores this property as a character vector.

Data Types: char | string

NumInputs — Number of inputs

2 (default)

Number of inputs of the layer. This layer has two inputs.

Data Types: double

InputNames — Input names

{'scores' 'boxDeltas'} (default)

Input names of the layer. This layer has two inputs, named 'scores' and 'boxDeltas'.

Data Types: cell

NumOutputs — Number of outputs

1 (default)

This property is read-only.

Number of outputs from the layer, returned as 1. This layer has a single output only.

Data Types: double

OutputNames — Output names

{'out'} (default)

This property is read-only.

Output names, returned as {'out'}. This layer has a single output only.

Data Types: cell

Examples

Create Region Proposal Layer

Define three square anchor boxes for the region proposal layer.

anchorBoxes = [
    16 16
    64 64
    128 128
    ];

Create a region proposal layer with the name 'region_proposal'.

regionProposal = regionProposalLayer(anchorBoxes,'Name','region_proposal');

References

[1] Ren, S., K. He, R. Girshick, and J. Sun. "Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks." Advances in Neural Information Processing Systems. Vol. 28, 2015.

Version History

Introduced in R2018b

R2024b: Not Recommended

Starting in R2024b, R-CNN object detectors are no longer recommended. Instead, use a different type of object detector, such as a yoloxObjectDetector or yolov4ObjectDetector object. These object detectors are faster than R-CNN object detectors. For more information, see Choose an Object Detector.