Read different scanned invoice

[mohammedamaan]

Hi. I have a scenario where I need to read different invoice. These invoice are scanned and are available in a folder. The invoice structure will be different for different invoice but I need to capture similar data like invoice number, date, amount, etc. How can I achieve this. If it can be explained as a step would be appreciated.

[anil5 (Anil Kumar Bandam)]

Hi @mohammedamaan,

You can use read PDF text with OCR activity, and see if the output is readable and because the invoices are scanned the data won’t be proper.

If the data is proper you can use regular expressions to extract the data.
If the data is not proper, use computer vision activities.

AI Computer Vision is Now Available for Preview

@Athrix05, Was successfully able to install computer vision activities by following the below steps Uninstalled UiPath.UIAutomation.Activities. Installed again Uipath UIAutomation Activities v19.2.0 Added the beta feed https://www.myget.org/F/uipath in Package Manager > Settings >User defined packag After adding the above feed in User Defined Package, the feed name will be displayed under All Packages Go to the Feed name, that was recently created and check IncludePrerelease and install U…

[mohammedamaan]

Hi @anil5 ,
As you said the data is not proper and I have no clue on how to use Computer vision. Can you just give me a short description how I can use it.

[anil5 (Anil Kumar Bandam)]

Hi @mohammedamaan,

Go through the provided link and install the activities.

After installation , use CV screen scope first indicate on screen from where you want to extract the values.

Inside screen scope, use CV get text to extract the values and using CV activities the data extracted is accurate.

To know more about CV activities, go through the activities guide for computer vision.

[mohammedamaan]

Ok. Thanks mate.

[mohammedamaan]

Hi @anil5,
I tried using the CV automation and seems to be good. But the issue i am facing how can I read data from scanned invoice for which the formats are not same. Also is there any way where we can read without opening the pdf file.

[mohammedamaan]

The accuracy is not 100% while reading pdf with microsoft OCR. How can i make it more accurate.

[Jeven_Delacruz]

If you’re dealing with image try to convert the PDF scanned docs to image(s) and use API Integration OCR

You can try using Microsoft Azure Computer Vision or Google Vision.
For Microsoft use the Microsoft Vision Activities > Handwritten Text (Mode as Printed)
Microsoft Vision requires the following:

• Service URL “https://[region].api.cognitive.microsoft.com”
• Subscription Key

Hope it helps

Regards,
Jeven

[pranaysai46]

hello @mohammedamaan

am a learner ,can you please attach the work flow how to get the invoices from different images

thanks.