detect

Detect objects using YOLO v2 object detector

Syntax

bboxes = detect(detector,I)

[bboxes,scores] = detect(detector,I)

[___,labels] = detect(detector,I)

[___,labels,info] = detect(detector,I)

detectionResults = detect(detector,ds)

[___] = detect(___,roi)

[___] = detect(___,Name=Value)

Description

bboxes = detect(detector,I) detects objects within a single image or an array of images, I, using you only look once version 2 (YOLO v2) object detector. The input size of the image must be greater than or equal to the network input size of the pretrained detector. The locations of objects detected are returned as a set of bounding boxes.

Note

The use of a CUDA^®-enabled NVIDIA^® GPU is highly recommended. The GPU reduces computation time significantly. Usage of the GPU requires Parallel Computing Toolbox™. For information about the supported compute capabilities, see GPU Computing Requirements (Parallel Computing Toolbox).

[bboxes,scores] = detect(detector,I) also returns the class-specific confidence scores for each bounding box.

[___,labels] = detect(detector,I) returns a categorical array of labels assigned to the bounding boxes in addition to the output arguments from the previous syntax. The labels used for object classes are defined during training using the trainYOLOv2ObjectDetector function.

[___,labels,info] = detect(detector,I) also returns information about the class probabilities and objectness scores for each detection.

detectionResults = detect(detector,ds) detects objects within all the images returned by the read function of the input datastore.

[___] = detect(___,roi) detects objects within the rectangular search region specified by roi. Use output arguments from any of the previous syntaxes. Specify input arguments from any of the previous syntaxes.

[___] = detect(___,Name=Value) specifies options using one or more name-value arguments in addition to any combination of arguments from previous syntaxes. For example, Threshold=0.75 sets the threshold to remove detections to 0.75.

Examples

Detect Vehicles Using YOLO v2 Object Detection Network

Load a YOLO v2 object detector pretrained to detect vehicles.

vehicleDetector = load('yolov2VehicleDetector.mat','detector');
detector = vehicleDetector.detector;

Read a test image into the workspace.

I = imread('highway.png');

Display the input test image.

imshow(I);

Run the pretrained YOLO v2 object detector on the test image. Inspect the results for vehicle detection. The labels are derived from the ClassNames property of the detector.

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

bboxes = 1×4

    78    81    64    63

scores = single

0.6224

labels = categorical
     vehicle 

Annotate the image with the bounding boxes for the detections.

if ~isempty(bboxes)
    detectedI = insertObjectAnnotation(I,'rectangle',bboxes,cellstr(labels));
end
figure
imshow(detectedI)

Input Arguments

detector — YOLO v2 object detector

yolov2ObjectDetector object

YOLO v2 object detector, specified as a yolov2ObjectDetector object. To create this object, call the trainYOLOv2ObjectDetector function with the training data as input.

I — Input image

H-by-W-by-C-by-B numeric array of images

Input image, specified as an H-by-W-by-C-by-B numeric array of images. Images must be real, nonsparse, grayscale or RGB image.

• H — Height in pixels.

• W — Width in pixels

• C — The channel size in each image must be equal to the network's input channel size. For example, for grayscale images, C must be equal to 1. For RGB color images, it must be equal to 3.

• B — Number of images in the array.

The detector is sensitive to the range of the input image. Therefore, ensure that the input image range is similar to the range of the images used to train the detector. For example, if the detector was trained on uint8 images, rescale this input image to the range [0, 255] by using the im2uint8 or rescale function. The size of this input image should be comparable to the sizes of the images used in training. If these sizes are very different, the detector has difficulty detecting objects because the scale of the objects in the input image differs from the scale of the objects the detector was trained to identify. Consider whether you used the SmallestImageDimension property during training to modify the size of training images.

Data Types: uint8 | uint16 | int16 | double | single | logical

ds — Datastore

datastore object

Datastore, specified as a datastore object containing a collection of images. Each image must be a grayscale, RGB, or multichannel image. The function processes only the first column of the datastore, which must contain images and must be cell arrays or tables with multiple columns.

roi — Search region of interest

[x y width height] vector

Search region of interest, specified as a four-element vector of the form [x y width height]. The vector specifies the upper left corner and size of a region in pixels.

Name-Value Arguments

Specify optional pairs of arguments as Name1=Value1,...,NameN=ValueN, where Name is the argument name and Value is the corresponding value. Name-value arguments must appear after other arguments, but the order of the pairs does not matter.

Before R2021a, use commas to separate each name and value, and enclose Name in quotes.

Example: Threshold=0.25 sets the threshold to remove detections to 0.25.

Threshold — Detection threshold

0.5 (default) | scalar in the range [0, 1]

Detection threshold, specified as a scalar in the range [0, 1]. Detections that have scores less than this threshold value are removed. To reduce false positives, increase this value.

SelectStrongest — Select strongest bounding box

true (default) | false

Select the strongest bounding box for each detected object, specified as true or false.

• true — Returns the strongest bounding box per object. The method calls the selectStrongestBboxMulticlass function, which uses nonmaximal suppression to eliminate overlapping bounding boxes based on their confidence scores.

  By default, the selectStrongestBboxMulticlass function is called as follows

   selectStrongestBboxMulticlass(bbox,scores,...
                                 RatioType="Union",...
                                 OverlapThreshold=0.5);

• false — Return all the detected bounding boxes. You can then write your own custom method to eliminate overlapping bounding boxes.

MinSize — Minimum region size

[1 1] (default) | vector of the form [height width]

Minimum region size, specified as a vector of the form [height width]. Units are in pixels. The minimum region size defines the size of the smallest region containing the object.

By default, MinSize is 1-by-1.

MaxSize — Maximum region size

size(I) (default) | vector of the form [height width]

Maximum region size, specified as a vector of the form [height width]. Units are in pixels. The maximum region size defines the size of the largest region containing the object.

By default, MaxSize is set to the height and width of the input image, I. To reduce computation time, set this value to the known maximum region size for the objects that can be detected in the input test image.

MiniBatchSize — Minimum batch size

128 (default) | scalar

Minimum batch size, specified as a scalar value. Use the MiniBatchSize to process a large collection of image. Images are grouped into minibatches and processed as a batch to improve computation efficiency. Increase the minibatch size to decrease processing time. Decrease the size to use less memory.

ExecutionEnvironment — Hardware resource

"auto" (default) | "gpu" | "cpu"

Hardware resource on which to run the detector, specified as "auto", "gpu", or "cpu".

• "auto" — Use a GPU if it is available. Otherwise, use the CPU.

• "gpu" — Use the GPU. To use a GPU, you must have Parallel Computing Toolbox and a CUDA-enabled NVIDIA GPU. If a suitable GPU is not available, the function returns an error. For information about the supported compute capabilities, see GPU Computing Requirements (Parallel Computing Toolbox).

• "cpu" — Use the CPU.

Acceleration — Performance optimization

"auto" (default) | "mex" | "none"

Performance optimization, specified one of the following:

• "auto" — Automatically apply a number of optimizations suitable for the input network and hardware resource.

• "mex" — Compile and execute a MEX function. This option is available when using a GPU only. Using a GPU requires Parallel Computing Toolbox and a CUDA enabled NVIDIA GPU. If Parallel Computing Toolbox or a suitable GPU is not available, then the function returns an error. For information about the supported compute capabilities, see GPU Computing Requirements (Parallel Computing Toolbox).

• "none" — Disable all acceleration.

The default option is "auto". If "auto" is specified, MATLAB^® applies a number of compatible optimizations. If you use the "auto" option, MATLAB does not ever generate a MEX function.

Using the Acceleration options "auto" and "mex" can offer performance benefits, but at the expense of an increased initial run time. Subsequent calls with compatible parameters are faster. Use performance optimization when you plan to call the function multiple times using new input data.

The "mex" option generates and executes a MEX function based on the network and parameters used in the function call. You can have several MEX functions associated with a single network at one time. Clearing the network variable also clears any MEX functions associated with that network.

The "mex" option is only available for input data specified as a numeric array, cell array of numeric arrays, table, or image datastore. No other types of datastore support the "mex" option.

The "mex" option is only available when you are using a GPU. You must also have a C/C++ compiler installed. For setup instructions, see MEX Setup (GPU Coder).

"mex" acceleration does not support all layers. For a list of supported layers, see Supported Layers (GPU Coder).

Output Arguments

bboxes — Location of objects detected

M-by-4 matrix | B-by-1 cell array

Location of objects detected within the input image or images, returned as an M-by-4 matrix or a B-by-1 cell array. M is the number of bounding boxes in an image, and B is the number of M-by-4 matrices when the input contains an array of images.

Each row of bboxes contains a four-element vector of the form [x y width height]. This vector specifies the upper left corner and size of that corresponding bounding box in pixels.

scores — Detection scores

M-by-1 numeric vector | T-by-1 cell array

Detection confidence scores for each bounding box, returned as one of these options:

• M-by-1 numeric vector — The input is a single test image. M is the number of bounding boxes detected in the image.

• B-by-1 cell array — The input is a batch of test images, where B is the number of test images in the batch. Each cell in the array contains an M-element row vector, where each element indicates the detection score for a bounding box in the corresponding image.

A higher score indicates higher confidence in the detection. The confidence score for each detection is a product of the corresponding objectness score and maximum class probability. The objectness score is the probability that the object in the bounding box belongs to a class in the image. The maximum class probability is the largest probability that a detected object in the bounding box belongs to a particular class.

labels — Labels for bounding boxes

M-by-1 categorical vector | T-by-1 cell array

Labels for bounding boxes, returned as one of these options:

• M-by-1 categorical vector if the input is a single test image.

• T-by-1 cell array if the input is an array of test images. T is the number of test images in the array. Each cell in the array contains a M-by-1 categorical vector containing the names of the object classes.

M is the number of bounding boxes detected in an image.

detectionResults — Detection results

3-column table

Detection results, returned as a 3-column table with variable names, Boxes, Scores, and Labels. The Boxes column contains M-by-4 matrices, of M bounding boxes for the objects found in the image. Each row contains a bounding box as a 4-element vector in the format [x,y,width,height]. The format specifies the upper-left corner location and size in pixels of the bounding box in the corresponding image.

info — Class probabilities and objectness scores

structure array

Class probabilities and objectness scores of the detections, returned as a structure array with these fields.

• ClassProbabilities — Class probabilities for each of the detections, returned as a B-by-1 cell array. B is the number of images in the input batch of images, I. Each cell in the array contains the class probabilities as an M-by-N numeric matrix. M is the number of bounding boxes and N is the number of classes. Each class probability is a numeric scalar, indicating the probability that the detected object in the bounding box belongs to a class in the image.

• ObjectnessScores — Objectness scores for each of the detections, returned as a B-by-1 cell array. B is the number of images in the input batch of images, I. Each cell in the array contains the objectness score for each bounding box as an M-by-1 numeric vector. M is the number of bounding boxes. Each objectness score is a numeric scalar, indicating the probability that the bounding box contains an object belonging to one of the classes in the image.

More About

Data Preprocessing

By default, the detect function preprocesses the test image for object detection by:

• Resizing it to a nearest possible image size used for training the YOLO v2 network. The function determines the nearest possible image size from the TrainingImageSize property of the yolov2ObjectDetector object.

• Normalizing its pixel values to lie in same range as that of the images used to train the YOLO v2 object detector. For example, if the detector was trained on uint8 images, the test image must also have pixel values in the range [0, 255]. Otherwise, use the im2uint8 or rescale function to rescale the pixel values in the test image.

Extended Capabilities

C/C++ Code Generation

Generate C and C++ code using MATLAB® Coder™.

Usage notes and limitations:

• The roi argument to the detect method must be a code generation constant (coder.const()) and a 1x4 vector.

• Only the Threshold, SelectStrongest, MinSize, and MaxSize name-value pairs for detect are supported.

GPU Code Generation

Generate CUDA® code for NVIDIA® GPUs using GPU Coder™.

Usage notes and limitations:

• The roi argument to the detect method must be a codegen constant (coder.const()) and a 1x4 vector.

• Only the Threshold, SelectStrongest, MinSize, MaxSize, and MiniBatchSize Name-Value pairs are supported.

• The height, width, channel, and batch size of the input image must be fixed size.

• The minimum batch size value passed to detect method must be fixed size.

GPU Arrays
Accelerate code by running on a graphics processing unit (GPU) using Parallel Computing Toolbox™.

Version History

Introduced in R2019a

R2024a: Option to return class probabilities and objectness scores

Specify the info output argument to return information about the class probability and objectness score for each detection.