IssueTracker

Business Impact: Bangle text and numbers are not getting extracted by Cloud Vision.

Issue summary: - The customer has the source type PDF & Image and in which content type is in Bangla language.

The customer aim is to fetch all the text from the given source.
Pain of the customer is that some images are able to perform the right extraction but not all. Gray-scaling the images is also not helpful.

Customer already overlooked the following links [a][b]:

[a] https://cloud.google.com/vision/docs/drag-and-drop
[b] https://cloud.google.com/vision/docs/supported-files
The customer already checked the image detection parameter in the python code as well:
response = client.text_detection(
image=image,
image_context={"language_hints": ["bn"]}, # Bengali)


As per the current update, our engineering teams are planning on releasing a new model that has a much better performance on Bangla text on "builtin/weekly" end of this quarter/beginning next quarter.