Activating AI objectdector when picture is taken with the camera

(0) Share

Report

Posted on by Highboy

1,183 Super User 2024 Season 1

Hello

I have trained an AI objectdetector, and woulld like to use it without manually pressing "detect"

I have a media/camera and a button that takes a picture when the button is pressed.

I want to use that picture in the AI, without any furter user input.

Is that possible?

Categories:

AI Builder

I have the same question (0)

All responses (8)

Answers (0)

billba Microsoft Employee on at

Like (0)

Report
This is possible using an experimental feature that we will be talking more about at Ignite next month.

1. In Settings/Upcoming features make sure "AI models as data sources" is turned on.
2. Select the "data" icon on the far left
3. Select "Add data"
4. In "Select a data source" you should now see an "AI models" option. Select that.
5. Your object detection model should show up on this list. Select it.

Now that your model has been added to the app, you can access it from any control using PowerFx. It has a single method called "Predict" which has different inputs and outputs depending on the model type. For Object Detection the input is an image and the output is a table called "results", each of which contains a tagId, confidence, and boundingBox.

For instance you might do something like this:
Button.OnSelect = Set(tagId, First('Your Model'.Predict(Camera1.Photo).responseV2.predictionOutput.results).tagId) Label.Text = tagId

In the near future this API will change and the new expression would look like:
Button.OnSelect = Set(tagId, First('Your Model'.Predict(Camera1.Photo).results).tagId)

I hope this helps!

Was this reply helpful? Yes No
Highboy 1,183 Super User 2024 Season 1 on at

Like (0)

Report

Thanks, I will have to test it in the morning.
It's a very well written answer.👍

Was this reply helpful? Yes No
rohitK264 4 on at

Like (0)

Report

Hi,

I am also facing the similar issue where I want to feed image taken from camera control in powerapps to AI Object Detector.

Have you got anything on this yet ?

Was this reply helpful? Yes No
billba Microsoft Employee on at

Like (0)

Report

Did you try the technique I cited above? That’s the way to do this.

Was this reply helpful? Yes No
vkatariya 5 on at

Like (0)

Report

Hi,

I am also facing the similar issue where I want to feed image taken from camera control in powerapps to AI Object Detector.

Have you got anything on this yet ?

Was this reply helpful? Yes No
billba Microsoft Employee on at

Like (0)

Report

Yes, you can use the technique described above.

Was this reply helpful? Yes No
rohitK264 4 on at

Like (0)

Report

Yes I tried but it seems to be not working for me as I did not want to detect the objects but instead faces. So object detection model did not work for me. Also I did not get to detect object without having user to click on Capture button.

Was this reply helpful? Yes No
billba Microsoft Employee on at

Like (0)

Report

Face detection models are somewhat different animals. Here's a thread that talks about using a different kind model. Keep in mind that you pay per model invocation, so if you want to run a model continuously (rather than on a button press) you will end up paying more, potentially a lot more. Be very careful how often you poll.

Was this reply helpful? Yes No