Multiple Object detection how to assign labels

(0) Share

Report

Posted on by JorgeLeigh

Hi everyone!

I have a problem.

I have deployed an Object Detection Model that can predict 3 different classes.
Everything works just fine. The problem is that when I have let´s say more than 3 objects in the picture, I see the different prediction bounding boxes in the picture, but when I want to evaluate the prediction confidence, I do not know to which object refers to. In order to solve it, I would need a number next to the label on the picture and on the prediction list, so I can link each object to its own prediction.

I hope I made myself clear.

And I wish there is a solution.

All the best!

Jorge

Categories:

AI Builder

I have the same question (0)

All responses (4)

Answers (1)

Sort by

Ashbhatia Moderator on at

Like (0)

Report
Copy link

Link copied!

Hi Jorge,

Interesting problem and ask. Are the objects too close to each other that you can't identify the object and its prediction? Would it be possible to send a screen shot?

Was this reply helpful? Yes No
JorgeLeigh 18 on at

Like (1)

Report
Copy link

Link copied!

Hi Ashbhatia! Let me sent you a screenshot.
As you can see, I have two visible labels, and in outcome table I have the label prediction plus its probability. But how can I know to which bounding box the first 10-30% prediction with confidence 0.8358768 belongs to? The same with the below 40-70%, I do not know to which image they represent the probability. Basically I am missing the link between the prediction and the picture. I went through the documentation but I could not find any related information.
I hope I made myself clear and looking forward to your comments.
Thanks in advance!

Jorge

Problem identifyi...

Was this reply helpful? Yes No
Verified answer

CedrickB Microsoft Employee on at

Like (0)

Report
Copy link

Link copied!
Here is what you can do

- First set the property ShowBoundingBoxes of the ObjectDetector component to false, so we hide the default bounding boxes.

- Then, add the BoundingBox column to your result table and set the property Visible of this column to false.

- Last, insert a Rectangle component that will serve as our dynamic bounding box display and set the properties to

Height=DataTable1.Selected.BoundingBox.Height*(ObjectDetector1.Height)
Width=DataTable1.Selected.BoundingBox.Width*ObjectDetector1.Width
X=DataTable1.Selected.BoundingBox.Left*ObjectDetector1.Width+ObjectDetector1.X
Y=DataTable1.Selected.BoundingBox.Top*ObjectDetector1.Height+ObjectDetector1.Y

With DataTable1 being you result table.

Now each time you will select a row in your table, the rectangle will appropriately show up on the image.

Was this reply helpful? Yes No
JorgeLeigh 18 on at

Like (0)

Report
Copy link

Link copied!

@CedrickB thank you so much!
It nicely worked!
I have marked it as a solution :).
All the best!

Was this reply helpful? Yes No