How to store value of tagged object - Object Detection Model

(0) Share

Report

Posted on by AVFC

147

Hello

Bit of background - I'm trying to create a meter reading application for old analog electricity meters. I've created an object detection model and trained it on the types of meters it might come accross and tagged the images with a meter reading location. I'm aware that object detection can't do text recognition at the same time.

My question is how do i store the value of my tagged Meter reading area into a variable which i could then scan using the text recognition AI model and then store that value in a SP list or Dataverse table? The best i've managed in my power automate flow is sending myself an email as a test with the detected object name of "Meter Reading" which isn't a good test anway..

I'm very new to power apps so any help would be most appreciated!

Categories:

AI Builder

All responses (14)

Answers (1)

Sort by

CedrickB Microsoft Employee on at

Like
a
(0)

Report
Copy link

Link copied!

What you are trying to do would require to crop the image based on bounding boxes found which is not feasible as far as I know.

Though, this thread may be an inspiration on how to approach your problem a different way

Solved: Re: The object detector does not work correctly wi... - Power Platform Community (microsoft.com)
AVFC 147 on at

Like
a
(0)

Report
Copy link

Link copied!

Thanks @CedrickB for the reply but the issue with using just text recogniser is that it picks up all the text on the front of the meter and i want just the meter reading value. One of your colleagues mentioned the below on another thread about object detection & text recognition at the same time as below, guess i'm trying to achieve that and maybe i need to use a canvas app rather then trying to send myself the value in an email?
CedrickB Microsoft Employee on at

Like
a
(0)

Report
Copy link

Link copied!

I just discussed with Joe and we both believe that the above solution he posted is not relevant to your problem.

Unfortunatelly, I don't see other option that the link provided above.
AVFC 147 on at

Like
a
(0)

Report
Copy link

Link copied!

Ok thanks for trying. I will look into your suggestion and see if it gives any ideas..
AVFC 147 on at

Like
a
(0)

Report
Copy link

Link copied!

Hi again, can i ask if i was to use text recognition, say i had an image with 4 lines of text on it and the for example the 2nd line is my meter reading value. How could i only capture the 2nd line and place that into a variable using for each text in current line?
CedrickB Microsoft Employee on at

Like
a
(0)

Report
Copy link

Link copied!

You can create an integer variable like this before the "Apply to each"

Inside the "Apply to each"
Add a condition to check when the variable is equal to 2 and capture the line when it is true.

And increment the variable at the end of the "Apply to Each" loop
AVFC 147 on at

Like
a
(0)

Report
Copy link

Link copied!

Thanks for all your help! Could you explain more the loop part, Would i use the outputs below for text recogniser before doing the condition:
CedrickB Microsoft Employee on at

Like
a
(0)

Report
Copy link

Link copied!

The condition should be inside the inner loop
AVFC 147 on at

Like
a
(0)

Report
Copy link

Link copied!

Thanks!
AVFC 147 on at

Like
a
(0)

Report
Copy link

Link copied!

Hi @CedrickB I've tried the below but when testing with pdf with 3 lines of text i'm getting line 3 & 4 on the email when i only want line 2. Can you please advise where i'm going wrong?

If i'm taking too much of your time then no worries!