Tesseract install

[tamasv]

I want to use pytesseract in a notebook. I have added pytesseract to my environment. I can import it.

However it does not work, I get an error when trying to run:

pytesseract.image_to_string()

Error:

"tesseract is not installed or it's not in your PATH. See README file for more information."

In my local machine I had to install the Tesseract at UB Mannheim windows exe, and I had to provide the path to the installed file.

I am guessing Fabric is missing this Tesseract at UB Mannheim installation.

Anyone figured out installing tesseract?

[v-prasare]

@tamasv  As we haven’t heard back from you, we wanted to kindly follow up to check if the solution provided by our super users for your issue worked? or let us know if you need any further assistance here?

@cmaneu, @nilendraFabric, Thanks for your promt response here.

Thanks,

Prashanth Are

MS Fabric community support

If this post helps, then please consider Accept it as the solution to help the other members find it more quickly and give Kudos if helped you resolve your query

[cmaneu · Accepted solution]

Hello @tamasv,
@nilendraFabric is right. PyTesseract relies on Tesseract binary and libraries. They're not part of the Fabric environment, and there is no easy way to download them. Depending on your use case, you may have several options.

Use EasyOCR

@nilendraFabric suggestion was good. This library works within Fabric. Here is a sample code.

%pip install easyocr
import easyocr
reader = easyocr.Reader(['en'])
result = reader.readtext("/lakehouse/default/Files/screenshot.png")
for detection in result:
    print(detection[1])

In my (small) experience, tesseract provides better results than easyOCR, so please check your use cases.

Use Azure Services

Azure Vision AI Services provides several ML models to extract both printed and handwritter text. With Document intelligence, you can even extract structured information - for example parsing an image of an invoice and automatically get each line item.

Tesseract - manually build and reference
If you really want to use Tesseract, Technically, you could hand-install tesseract packages, this will involve manually downloading (deb) packages and untar them, but that would be quite time-consuming with all the chain of dependency. You could also compile it yourself to have a single exe with all the dependencies linked (someone on the Internet may have done that alreay).

Hope this helps! 

[tamasv]

Hello!

Thanks for the info provided. EasyOCR does work. Just the project is not built around that library. As our experience shows that easyocr is a bit worse for what we need it. Manual build is not my expretise, but maybe we will look into it. Also on the long run Azure Vision could be a potential candidate.

I accept this solution as it opened up some options to do the project in Azure ecosystem.

Thanks

[nilendraFabric]

Hello @tamasv 

You are right.

Fabric does not natively support the Tesseract OCR engine or its Linux-based binaries, which are required for pytesseract to function.

Did you tried you using EasyOCR ,does not require external binaries.

hope this is helpful. 
please accept the answer and give kudos if this helps