How to put a bounding box for image detection using SVM model exported from Classification Learner?

Hello, I have a classifier trained with Classification Learner APP, 200-word-bagOfFeature per image was used as trained data.

   PoseScrewClassifier = 

struct with fields:

           predictFcn: @(x)exportableClassifier.predictFcn(predictorExtractionFcn(x))
    RequiredVariables: {1×200 cell}
    ClassificationSVM: [1×1 ClassificationSVM]
                About: 'This struct is a trained classifier exported from Classifi…'
         HowToPredict: 'To make predictions on a new table, T, use: …'

I have used it to predict new data, T. T is a table summarize new testing images with 200-word-bagOfFeature per image.

if true
[label, score] = predict(PoseScrewClassifier.ClassificationSVM,T);
end

I want to put a bounding box around the object which the classifier thinks it belongs to certain category. I want to do this because I have a lot of testing images not only have category 1, but also have category 2 in the same image. This is what I did...

if true
  [bboxes, score] = predict(PoseScrewClassifier.ClassificationSVM,T);
  box = insertObjectAnnotation(imageT,'rectangle',bboxes,score);
  figure, imshow(box);
end

And this is the Error I got:

if true
Error using insertObjectAnnotation
Expected input number 3, POSITION, to be one of these types:
numeric
Instead its type was categorical.
Error in insertObjectAnnotation
Error in insertObjectAnnotation  
end

I do not quite understand. Is it because of the nature of bagoffeature which is looking at the entire image, so there will not be a bounding box at all? Or some other reasons?

And what are the possible solutions to get a bounding box? A sliding window?

Someone please point me a direction. Greatly appreciate!