yolov4ObjectDetector - Detect objects using YOLO v4 object detector - MATLAB (original) (raw)

Detect objects using YOLO v4 object detector

Since R2022a

Description

The yolov4ObjectDetector object creates a you only look once version 4 (YOLO v4) one-stage object detector for detecting objects in an image. Using this object, you can:

Create a pretrained YOLO v4 object detector by using YOLO v4 deep learning networks trained on COCO dataset.
Create a custom YOLO v4 object detector by using any pretrained or untrained YOLO v4 deep learning network.

Creation

Syntax

Description

Pretrained YOLO v4 Object Detector

`detector` = yolov4ObjectDetector([name](#mw%5F8537ba41-666d-4029-aa87-4edb86cedda5)) creates a pretrained YOLO v4 object detector by using YOLO v4 deep learning networks trained on a COCO dataset.

example

Custom YOLO v4 Object Detector

`detector` = yolov4ObjectDetector([name](#mw%5F8537ba41-666d-4029-aa87-4edb86cedda5),[classes](#mw%5F004cf8ab-48db-439a-ad7f-9f2fa65ef293),[aboxes](#mw%5Fdc4b928a-a225-4a55-8b0f-420e9e5945f3)) creates a pretrained YOLO v4 object detector and configures it to perform transfer learning using a specified set of object classes and anchor boxes. For optimal results, you must train the detector on new training images before performing detection. Use thetrainYOLOv4ObjectDetector function for training the detector.

`detector` = yolov4ObjectDetector([net](#mw%5Fcf65ac24-5375-4dfe-9400-370ac0909a34),[classes](#mw%5F004cf8ab-48db-439a-ad7f-9f2fa65ef293),[aboxes](#mw%5Fdc4b928a-a225-4a55-8b0f-420e9e5945f3)) creates an object detector by using the deep learning networknet.

If net is a pretrained YOLO v4 deep learning network, the function creates a pretrained YOLO v4 object detector. The classes and aboxes are values used for training the network.

If net is an untrained YOLO v4 deep learning network, the function creates a YOLO v4 object detector to use for training and inference.classes and aboxes specify the object classes and the anchor boxes, respectively, for training the YOLO v4 network.

Use the trainYOLOv4ObjectDetector function to train the network before performing object detection.

`detector` = yolov4ObjectDetector([baseNet](#mw%5F496e0a3a-f137-4b44-88b7-b8f10db1a876),[classes](#mw%5F004cf8ab-48db-439a-ad7f-9f2fa65ef293),[aboxes](#mw%5Fdc4b928a-a225-4a55-8b0f-420e9e5945f3),DetectionNetworkSource=[layer](#mw%5F8156f365-78ca-42bd-90b0-6fa27ed9b387)) creates a YOLO v4 object detector by adding detection heads to a base network,baseNet.

The function adds detection heads to the specified feature extraction layerslayer in the base network. To specify the names of the feature extraction layers, use the name-value argumentDetectionNetworkSource=layer.

If baseNet is a pretrained deep learning network, the function creates a YOLO v4 object detector and configures it to perform transfer learning with the specified object classes and anchor boxes.

If baseNet is an untrained deep learning network, the function creates a YOLO v4 object detector and configures it for object detection.classes and aboxes specify the object classes and the anchor boxes, respectively, for training the YOLO v4 network.

You must train the detector on a training dataset before performing object detection. Use the trainYOLOv4ObjectDetector function for training the detector.

example

`detector` = yolov4ObjectDetector(___,`Name=Value`) specifies one or more options using name-value arguments in addition to any combination of input arguments from previous syntaxes. Use this syntax to:

Modify the detection network sources in a YOLO v4 object detection network and train the network with different numbers of object classes, anchor boxes, or both. Specify the new detection network sources using theDetectionNetworkSource=[layer](#mw%5F8156f365-78ca-42bd-90b0-6fa27ed9b387) name-value argument.
Set the [InputSize](yolov4objectdetector.html#mw%5F0e561b1b-ca03-4f11-897e-78f2253f89f4) and [ModelName](yolov4objectdetector.html#mw%5Fa8bf28f7-40bb-45fb-b68c-5b0cd224bcf4) properties of the object detector. For example,InputSize=[224 224 3] sets the size of the images used for training to [224 224 3].

Note

To use the pretrained YOLO v4 object detection networks trained on COCO dataset, you must install the Computer Vision Toolbox™ Model for YOLO v4 Object Detection. You can download and install the Computer Vision Toolbox Model for YOLO v4 Object Detection from Add-On Explorer. For more information about installing add-ons, seeGet and Manage Add-Ons. To run this function, you will require the Deep Learning Toolbox™.

Input Arguments

expand all

Name of the pretrained YOLO v4 deep learning network, specified as one of these:

"csp-darknet53-coco" — A pretrained YOLO v4 deep learning network created using CSP-DarkNet-53 as the base network and trained on COCO dataset.
"tiny-yolov4-coco" — A pretrained YOLO v4 deep learning network created using a small base network and trained on COCO dataset.

Data Types: char | string

Names of object classes for training the detector, specified as a string vector, cell array of character vectors, or categorical vector. This argument sets theClassNames property of the yolov4ObjectDetector object.

Data Types: char | string | categorical

Anchor boxes for training the detector, specified as an _N_-by-1 column vector cell array or a 1-by-N row vector cell array.N is the number of output layers in the YOLO v4 deep learning network. Each cell contains an M_-by-2 matrix, where_M is the number of anchor boxes in that layer. Each cell can contain a different number of anchor boxes. Each row in the _M_-by-2 matrix denotes the size of an anchor box in the form [height _width_].

The first element in the cell array specifies the anchor boxes to associate with the first output layer, the second element in the cell array specifies the anchor boxes to associate with the second output layer, and so on. For accurate detection results, specify large anchor boxes for the first output layer and small anchor boxes for the last output layer. That is, the anchor box sizes must decrease for each output layer in the order in which the layers appear in the YOLO v4 deep learning network.

This argument sets the AnchorBoxes property of theyolov4ObjectDetector object.

Data Types: cell

YOLO v4 deep learning network, specified as a dlnetwork (Deep Learning Toolbox) object. The input network can be either an untrained or a pretrained deep learning network. The input network must have an image input layer.

Base network for creating the YOLO v4 deep learning network, specified as adlnetwork (Deep Learning Toolbox) object. The network can be either an untrained or a pretrained deep learning network. The input network must have an image input layer.

Names of the feature extraction layers in the base network, specified as a cell array of character vectors, or a string array. The function creates a YOLO v4 network by adding detection head layers to the output of the feature extraction layers in the base network.

In the pre-trained yolo v4 base network, you must connect the detection head layers to feature extraction layers where the input image feature map changes spatial dimension by downsampling factors of 4, 8, 16, and 32.

Use the analyzeNetwork (Deep Learning Toolbox) function to display the pre-trained YOLO v4 network architecture to obtain the information about the spatial dimensions of the feature maps.

Use this table to know about the default feature extraction layers and choose alternate feature extraction layers supported by the pre-trained YOLO v4 networks. Choose feature extraction layers using empirical evaluation based on the detection accuracy, training speed, and object size to detect.

To achieve higher detection accuracy, choose feature extraction layers with both higher and lower downsampling factors. Layers with lower downsampling factor capture small objects and layers with higher downsampling factor ensure a thorough feature refinement.
To achieve higher training and inference speeds, choose feature extraction layers with lower downsampling factor.

The number of detection network sources specified must be equal to the number of anchor boxes specified at the input.

Backbone	Number of Detection Heads	Default Feature Extraction Layers	Examples of Alternate Feature Extraction Layers	Remarks
Layer	Downsampling Factor	Layer	Downsampling Factor
tiny-yolov4-coco	2	leaky_28leaky_25	3216	leaky_17leaky_9	84	Select any two feature extraction layers with different downsampling factors preceding the leaky_28 layer to use as detection network sources.For example, specify the new detection network sources as layer = {'leaky_17','leaky_9'}
csp-darknet53-coco	3	mish_106mish_87mish_56	32168	mish_88mish_102	32	Select any three feature extraction layers with different downsampling factors preceding themish_106 layer to use as detection network sources.For example, specify the new detection network sources as layer = {'mish_57','mish_26','mish_13'}
mish_68mish_57	16
mish_37mish_26	8
mish_25mish_18mish_13	4

Data Types: char | string | cell

Properties

expand all

YOLO v4 deep learning network to use for object detection, stored as a dlnetwork (Deep Learning Toolbox) object.

Names of object classes to detect, stored as a categorical vector. You can set this property by using the input argument classes.

This property is read-only.

Set of anchor boxes, stored as a _N_-by-1 cell array.N is the number of output layers in the YOLO v4 deep learning network for which the anchor boxes are defined. Each element in the cell is a _M_-by-2 matrix. M denotes the number of anchor boxes. Each cell can contain a different number of anchor boxes. Each row in the _M_-by-2 matrix denotes the size of the anchor box in the form of [height _width_]. The first element in the cell array specifies the anchor boxes for the first output layer, the second element in the cell array specifies the anchor boxes for the second output layer, and so on.

You can set this property by using the input argumentaboxes.

This property is read-only.

Image size used for training, stored as a vector of form [height _width_] or [height width _channels_]. To set this property, specify it at object creation. The size of the training images must be a multiple of 32.

For example, detector = yolov4ObjectDetector(net,classes,aboxes,InputSize=[224 224 3]).

Name for the object detector, stored as a character vector or string scalar. To set this property, specify it at object creation.

For example,yolov4ObjectDetector(net,classes,aboxes,ModelName="customDetector") sets the name for the object detector to "customDetector".

This property is read-only.

Bounding box format for an object detector, stored as"axis-aligned" or "rotated". When thePredictedBoxType is "axis-aligned", the object detector will train and perform inference on only axis-aligned bounding boxes. If it is set to "rotated", the object detector will train and perform inference on only rotated bounding boxes. Set this property when you create the object.

Object Functions

detect	Detect objects using YOLO v4 object detector

Examples

collapse all

Specify the name of a pretrained YOLO v4 deep learning network.

name = "tiny-yolov4-coco";

Create YOLO v4 object detector by using the pretrained YOLO v4 network.

detector = yolov4ObjectDetector(name);

Display and inspect the properties of the YOLO v4 object detector.

yolov4ObjectDetector with properties:

    Network: [1×1 dlnetwork]
AnchorBoxes: {2×1 cell}
 ClassNames: {80×1 cell}
  InputSize: [416 416 3]
  ModelName: 'tiny-yolov4-coco'

Use analyzeNetwork to display the YOLO v4 network architecture and get information about the network layers.

analyzeNetwork(detector.Network)

Detect objects in an unknown image by using the pretrained YOLO v4 object detector.

img = imread("highway.png"); [bboxes,scores,labels] = detect(detector,img);

Display the detection results.

detectedImg = insertObjectAnnotation(img,"Rectangle",bboxes,labels); figure imshow(detectedImg)

This example shows how to create a YOLO v4 object detection network based on a pretrained ResNet-50 convolutional neural network.

Load a pretrained deep learning network to use as the base network. This example uses ResNet-50 pretrained network as the base network. For information about other available pretrained networks, see Pretrained Deep Neural Networks (Deep Learning Toolbox).

basenet = imagePretrainedNetwork("resnet50");

Use analyzeNetwork to display the architecture of the base network.

The first layer in the base network is the image input layer. Inspect the property of the image input layer in the base network.

ans = ImageInputLayer with properties:

                  Name: 'input_1'
             InputSize: [224 224 3]
    SplitComplexInputs: 0

Hyperparameters DataAugmentation: 'none' Normalization: 'zerocenter' NormalizationDimension: 'auto' Mean: [224×224×3 single]

To create a YOLO v4 deep learning network you must set the Normalization property of the ImageInputLayer in the base network to "none". Define an image input layer with the Normalization property set as "none" and other property values the same as those of the base network.

imageSize = basenet.Layers(1).InputSize; layerName = basenet.Layers(1).Name; newInputLayer = imageInputLayer(imageSize,Normalization="none",Name=layerName);

Replace the image input layer in the base network with the new input layer.

dlnet = replaceLayer(basenet,layerName,newInputLayer);

Specify the names of the feature extraction layers in the base network to use as the detection heads.

featureExtractionLayers = ["activation_22_relu","activation_40_relu"];

Specify the class names and anchor boxes to use for training the YOLO v4 deep learning network created using ResNet-50 as the base network.

classes = ["car","person"]; anchorBoxes = {[122,177;223,84;80,94];... [111,38;33,47;37,18]};

Create a YOLO v4 object detector by using the specified base network and the detection heads.

detector = yolov4ObjectDetector(dlnet,classes,anchorBoxes, ... DetectionNetworkSource=featureExtractionLayers);

Display and inspect the properties of the YOLO v4 object detector.

yolov4ObjectDetector with properties:

         Network: [1×1 dlnetwork]
     AnchorBoxes: {2×1 cell}
      ClassNames: {2×1 cell}
       InputSize: [224 224 3]
PredictedBoxType: 'axis-aligned'
       ModelName: ''

Use analyzeNetwork to display the YOLO v4 network architecture and get information about the network layers.

analyzeNetwork(detector.Network)

This example shows how to detect people using a pretrained YOLO v4 object detector.

Load the pretrained YOLO v4 detector that can detect 80 common object classes.

detector = yolov4ObjectDetector();

Read image to process.

I = imread("visionteam.jpg");

Run object detector.

[bboxes, scores, labels] = detect(detector, I, Threshold=0.4);

Select detections for the person class.

personBoxes = bboxes(labels=="person", :);

Display results.

detectedImg = insertObjectAnnotation(I, "Rectangle", personBoxes, "person"); figure imshow(detectedImg)

Figure contains an axes object. The hidden axes object contains an object of type image.

Extended Capabilities

expand all

The roi argument to the detect method must be a code generation constant (coder.const()) and a 1x4 vector.
Only the Threshold, SelectStrongest,MinSize, MaxSize, andMiniBatchSize name-value pairs for detect are supported.
The roi argument to the detect method must be a code generation constant (coder.const()) and a 1x4 vector.
Only the Threshold, SelectStrongest,MinSize, MaxSize, andMiniBatchSize name-value pairs for detect are supported.

For information about how to create a yolov4ObjectDetector object for code generation, see Load Pretrained Networks for Code Generation (MATLAB Coder).

Usage notes and limitations:

GPU Arrays is not supported for rotated rectangle bounding box inputs.

Version History

Introduced in R2022a

expand all

Starting in R2024b, you can modify the detection network sources of a YOLO v4 object detection network when you perform transfer learning. You can specify the new detection network sources using the name-value argumentDetectionNetworkSource=`layer`.

Specify new detection network sources for the tiny-yolov4-coco and csp-darknet53-coco pretrained YOLO v4 object detectors by using the syntax
detector = yolov4ObjectDetector(name,classes,aboxes,DetectionNetworkSource=layer);
Specify new detection network sources for a YOLO v4 object detection network by using the syntax
detector = yolov4ObjectDetector(net,classes,aboxes,DetectionNetworkSource=layer);

For an example, the below code modifies the detection network sources of the pretrainedtiny-yolov4-coco network to leaky_23 andleaky_15.

name = "tiny-yolov4-coco"; layer = ["leaky_23","leaky_15"]; detector = yolov4ObjectDetector(name,classes,aboxes,DetectionNetworkSource=layer);