Text Recognizer in API Builder.

[Anonymous]

I'm testing the Text Recognizer in API Builder. The app is able to detected the text / handwriting and draw a rectangular around the detected text. Now, I'm trying to extract this text to enable me to compare this extracted text to a list of text and do a match to return an different value. for instance, a number I wrote on the Fridge and I'll need the app to detect this number , compare it  against a list of furniture list and return the value  "Fridges". Please advise ?

[alana · Accepted solution]

Hello,

Text recognizer is great to extract printed and handwritten text from images, it will work perfectly with your scenario.

There are a lot of new features available soon for Text recognizer and all other AI Builder components so stay tuned.
One of the new features for Text recognizer is that you can select one of the rectangular bounding boxes around the extracted text lines in the image. The text from the selected bounding box can be retrieved using the property "SelectedText". So for your scenario you can use this property of the Text Recognizer component to match to the list of furniture you have by using the lookup function from Power Apps 

Thanks,

Alana

[Anonymous]

Hi Alana

Thanks for your help with this, the app is actually working as I've planned after using the SelectedText property. 

for moving the bounding rectangular box around the text that i need to retrieve - is it working now? i tired, but was not succesfful unless I'm doing something wrong.

Beside, the API itself works only on photo they were taken? what about if I want to use the camera instead, focus the Camera on the text or handwriting and retrieve the text ? is this feature available  now ? 

thanks again 

Jay 

[alana · Accepted solution]

Hello Jay,

Unfortunately we just surface the bounding boxes for the text our OCR technology was able to detect. You cannot draw a bounding box on the Text Recognizer component to define the text to extract.

Depending on the device, if you run a Power App with Text Recognizer, when you click on "+ New image", you will be proposed to take a picture, choose an image from your device's gallery and/or pick a file. You can propose a new idea if you want the component to use an input image as a source.

For your information, the Text Recognizer component extracts printed and handwritten text. This feature is already available.

Thanks,

Alana

[Anonymous]

Hello Alana,

I've tested the printed and handwriting and works very good. I was thinking if using the phone camera to act as a scanner just like scanning bar code or qr code. and Yes, I'd like to suggest the real time interaction of the source image in a dynamic and instant which in the background it can check against any list of text 

To point and auto focus to get the best clear text with the help of moving the detected bounding box by the OCR, not drawing a box rather. 

[Frank-AYES]

instead of click the + new image of the control box, we want to use the captured photo from the camera component, e.g. textreconginzer1.originalimage=camera1.photo. it seems to be not working.

[CedrickB]

Hi,

Here is how you should proceed

1. Add the Text recognition model as a datasource

2. In the "OnSelect" property of the camera control enter this formula

Set(TRResults, 'Text recognition'.Predict(Camera1.Photo).results)

3. Add any display component like a label and enter this in the "Text" property

First(First(TRResults.lines).lines).text

Here the formula retrieve the first line recognized but you can iterate with "FirstN" function

[Frank-AYES]

Hi, it works! Thanks a lot!
one more question, is there any way to mirror the picture taken by the camera stream?

[skylitedave]

Yes - put that code in the OnStream property of the camera...

Tip - Set Stream property to something like 15000 so camera1 "snaps" and image every 15 seconds 

Tip 2 - You can also set the default image of the AI text extractor control to Camer1.OnStream then hide the camera for automagic text extraction to display the image autocaptured by the camera

You are welcome 🙂

[SyedJaas]

Set(TRResults, 'Text recognition'.Predict(Camera1.Photo).results)

Seems results proprty is no more.I see only page property