How can fix the error 'Invalid calling syntax for the "predict" command.' when using the predict function?

I used fitcsvm to train INRIAPerson and when I was using the predict function to get the label of input images, I got an error which is 'Invalid calling syntax for the "predict" command'. But it worked in another code. It confused me and I really do not know how to fix it. Here is how I train the classifier:

    function TrainingSVM()
    %Input the training images and set image size.
    imgTrain  = imageDatastore('D:\TrainImage\INRIAPerson','IncludeSubfolders', true, 'LabelSource', 'foldernames');
    imageSize = [96, 48];
    Test = readimage(imgTrain,1);  
    Test = imresize(Test,imageSize);  
    [features, ~] = extractHOGFeatures(Test);  
    numImages = length(imgTrain.Files);  
    data = zeros(numImages,size(features,2),'single');
    for i = 1:numImages  
        Train = readimage(imgTrain,i);  
        Train = imresize(Train,imageSize);  
        data(i,:) = extractHOGFeatures(Train);  
    end
    trainLabels = imgTrain.Labels;     
    classifier = fitcsvm(data,trainLabels);  
    save classifier
    end

Here is the error code.

    function PedestrianTracking()

    %Load the pre-trained classifier
    classifier = 'classifier.mat';

    %Create object obj
    obj = setupSystemObjects(); 

    % Detect moving objects, and track them across video frames.
    while ~isDone(obj.reader)
        Frame = readFrame();
        frame = imresize(Frame, 0.5, 'Antialiasing',false);
        [centroids, bboxes, ~] = detectObjects(frame);
        Size = size(bboxes);
        %Find areas contain people and save them.
        Bboxes = [];
        Centroids = [];
        for i = 1:Size(1,1)
            %Get the target area
            detectImage = imcrop(frame, bboxes(i, :));
            %Resize the target area and do the classification with classifier
            DetectImage = imresize(detectImage, [96, 48]);
            feature = extractHOGFeatures(DetectImage);
            [predictIndex,~] = predict(classifier,feature);
            % If people are detected in the DetectImage then store it.
            if ( char(predictIndex) - 48 ) == 1
                Bboxes = [Bboxes; bboxes(i, :)];
                Centroids = [Centroids, centroids(i, :)]
            end 
        end   
    end

    function [centroids, bboxes, mask] = detectObjects(frame)
        % Detect foreground.
        mask = obj.detector.step(frame);
        % Apply morphological operations to remove noise and fill in holes.
        mask = imopen(mask, strel('rectangle', [3,3]));
        mask = imclose(mask, strel('rectangle', [15, 15]));
        mask = imfill(mask, 'holes');
        % Perform blob analysis to find connected components.
        [~, centroids, bboxes] = obj.blobAnalyser.step(mask);
    end

    function obj = setupSystemObjects()
        obj.reader = vision.VideoFileReader('D:D:\TestVideo\1.mp4');
        obj.detector = vision.ForegroundDetector('NumGaussians', 4, ...
            'NumTrainingFrames', 20, 'MinimumBackgroundRatio', 0.7);
        obj.blobAnalyser = vision.BlobAnalysis('BoundingBoxOutputPort', true, ...
            'AreaOutputPort', true, 'CentroidOutputPort', true, ...
            'MinimumBlobArea', 4000);
        obj.videoPlayer = vision.VideoPlayer('Position', [20, 400, 700, 400]);   
    end

    end

Answer 1

[Walter Roberson]

You have

      classifier = 'classifier.mat';

and you expect that to be enough to reference a saved classifier. It is not. You need to do something more like

     filename = 'classifier.mat'
     fileinfo = whos('-file', filename);
     varclasses = {fileinfo.class};
     matches = ~cellfun(@isempty, regexpi(varclasses, 'classification'));
     matchidx = find(matches, 1, 'first');
     if isempty(matchidx); error('File "%s" has no classifier stored in it', filename); end
     matchvar = fileinfo(matchidx).name;
     file_data = load(filename, matchvar);
     classifier = file_data.(matchvar);

... All of which would be much easier if only you already knew what the name of the variable was that had been used to store the classifier in the .mat file.

[BOFU LIU]

Yeah, I just found out that I only wrote down the file path but forgot to load the file. Thank you very much. But I met another problem with the same code after I load the classifier. It is said 'No valid system or dataset was specified.'

[Walter Roberson]

It appears that the code I posted can be abbreviated to:

filename = 'classifier.mat';
file_data = load(filename, 'classifier');
classifier = file_data.classifier;

You could try that and see if it makes a difference.