Re: Vision API - Text detection training

brford · 01-05-2022 03:59 AM

Hi there,

I'd like to train my Vision Project to improve document text detection for manuscript books. I couldn't find the solution anywhere. The current result is awful. The language is Portuguese.

Please advise. Thanks.

oakinlaja

Hello,

I hope I understand your concern. The issue seems to be about the Quality of the result observed.

Could you give more insights and details on what you observed? As per this doc[0], Portuguese is one of the Supported languages by the Cloud Vision API's text recognition feature. So, if the result is not what you expected, perhaps sharing more info about your use-case, your image/file and the output observed, will help to understand how we could help. You can share a screenshot as an example, while describing what you observed and what you expected.

[0]https://cloud.google.com/vision/docs/languages

oakinlaja

Please ensure that whatever Documents or Screenshots added to this thread does not include PII or SPII.

Thank you.

brford

Hi Oakinlaja,

Thanks for replying. I'm sending to you my code, image file and result. It does not include PII or SPII. Please advise.
https://1drv.ms/u/s!Ap4FzqZLgHXGpYlNTp8Y8Q6eP_7jiw?e=3tGSIx