Textract does not recognize image text

Sometimes in general while doing OCR (Optical Character Recognition) you need to pre-process the image, especially when you have some characters that not very defined.

In your case:

1. I would crop the image to "focus" on the non-white part
2. tilt the image in order to be as vertical as possible
3. increase the contrast by 20-30%

Therefore, as you can see from my result, it recognize everything :)

You could use a Lambda or the SageMaker Processing Python job with OpenCV or similar product in order to pre-process the images. Usually this is a best practice in order to obtain better results. Please check this blog: https://aws.amazon.com/it/blogs/machine-learning/process-text-and-images-in-pdf-documents-with-amazon-textract/ or this notebook: https://github.com/aws-samples/textract-visual-removal/blob/main/visual_removal_canny_edge_detector.ipynb

the output in Textract:

[if my response was useful, please click the "Accept" tick]

For This use case Consider using Amazon Rekognition - > https://docs.aws.amazon.com/rekognition/latest/dg/text-detection.html .