How to label multiple objects in object detection with different names?

Hi there!

I've a problem with labelling my objects in an image.

Let's have a look at the image:

This programme is detecting front/rear of the cars and a stop sign. I want the labels to say what they're looking at. For example: "Stop Sign Confidence: 1.0000", CarRear Confidence: 0.6446 and etc. As you may see, my programme adds the probability factors correctly. But there are still no strings/labelnames attached.

You can have a look at my code:

%%

% Read test image

testImage = imread('StopSignTest2.jpg');

% Detect stop signs

[bboxes,score,label] = detect(rcnn,testImage,'MiniBatchSize',128)

% Display detection results

label_str = cell(3,1);

conf_val = [score];

conf_lab = [label];

for ii=1:3

label_str{ii} = [ ' Confidence: ' num2str(conf_val(ii), '%0.4f')];

end

position = [bboxes];

outputImage = insertObjectAnnotation(testImage,'rectangle',position,label_str,...

'TextBoxOpacity',0.9,'FontSize',10);

figure

imshow(outputImage)

%%

I have NO clue in how to add strings in the label_str{ii} like the way I did with scores (num2str(conf_val(ii)).

Thanking you in advance!

[Sarannya Udhayakumar]

hello... how to detect multiple objects in RCNN???

pls rply

[Image Analyst]

Maybe there is only one object in your image.

Answer 1

[Image Analyst]

If insertObjectAnnotation() is not working, then use text().

[Banchon Sripanha]

Thanks for the response! Much appreciated.

I've fixed the problem by the following line code:

conf_val = [score];
conf_lab = [label];

for ii=1:3
    label_str{ii} = [sprintf('%s', conf_lab(ii)) ' Confidence: ' num2str(conf_val(ii), '%0.4f')];

Answer 2

[Quoc-Bao Ta]

Hello, How can i change the colors of bounding box?

[Image Analyst]

@Quoc-Bao Ta, use the AnnotationColor option of insertObjectAnnotation.

outputImage = insertObjectAnnotation(testImage,'rectangle',position,label_str,...
    'TextBoxOpacity',0.9,'FontSize',10, ...
    'LineWidth', 2, 'AnnotationColor', 'r');

Answer 3

[Onurcan]

hi, is it possible that you can share the code for multiple class detection. I've been trying but couldnt figure it out. I can only train one object but more than one doesnt work. please help