What OCR software i could opt for ?

A quick search shows that adobe has an online OCR service - Adobe Developer — PDF Services API — OCR PDF - Adobe Developers

​@hkingsbury Am not sure how to use it into FME, do you mean that instead of  tesseract.exe  I call PDF OCR ? Or use PythonCaller ? 
 

Highly appreciated if u got more on that or any example of workbench/ tnx 

Its a REST API so you’ll be able to call it via the HTTPCaller. I don’t have a specific example for this API. There is a generic tutorial here on how to use APIs via FME Tutorial: Getting Started with APIs – FME Support Center

Hi,

Have you considered a local vision model?  

​@antoine Thank you, am not aware of that, if you got any example / workflow would help a bit more to understand worth trying out. 

About the idea, you can read :
https://huggingface.co/blog/dvilasuero/how-to-analyze-images-with-ai
It can be set up locally if you have a good graphic card.

​@oliver.morris did a good presentation using Gemini about it but I cannot find the link. There are other services available through FME if your hardware is not good enough or if you have big volumes like GoogleVisionConnector, an Azure one, an AWS one, depending on your cloud provider.


 

It depends on the quality of the inputs so always best to try a few and see whcih model is best optimised for your inputs.

For more tricky work using external AI services like gemini is the way forward. But if you have a fairly modern machine and time on your hands (i.e. you can leave it to run over night) there are some good options.

Installing Ollama locally and using a small model like Paddle or GLM maybe a good starting point -https://ollama.com/MedAIBase/PaddleOCR-VL

https://ollama.com/library/glm-ocr

Hi ​@vimva679, building off the great suggestions already mentioned above, you can point an AI model at the image or file and have it extract and interpret the text.

• An example walkthrough of reading meter values from field photos: Getting Started with AI in FME: Image and Visual AI Tools
• Another great example with a real world use case from the Town of Qualicum Beach and Avineon Tensing going beyond just text extraction with a downloadable workspace: From Prototype to Production: Building Real-World AI and MCP Workflows

If you want to look into this further, you can refer to any of the AI connectors or tutorials in this article to get started: Getting Started with Any AI in FME.

Or if you’d rather stay with a dedicated OCR, a few connector options that may align with your IT requirements as suggested by ​@antoine :

• GoogleVisionConnector
• AzureAIVisionConnector
• RekognitionConnector (short text only)

Hope that points you somewhere useful!