IssueTracker

Hi there,

I am unable to retrieve and test your image as the bucket you provided does not seem to be available anymore. Would it be possible for you to share the original image(twoc.png) that you’ve used in the Cloud Vision API Explorer?

Alternatively, it would be important to verify if any of the best practices would apply to your image. For example, when using TEXT_DETECTION the OCR of Cloud Vision API requires more resolution to detect characters and therefore, a minimum image size of 1024 x 768 pixels is advised.

i attached example image file.
when i use other language in "Language Hints" like Arabic,English it show bad output.line will be break wrong.
but when i use Google doc , that will show correct output.i think Google doc use other mechanism of OCR or any preprocessing (ex:change size,layer recognition,stronger language analyzer or even any image processing and etc).
may be when you test this image it show good output but in other image (Asian language like Arabic) file it has bad output and user should edit text file.
how can i get output like Google doc with Google cloud text_detection.

[Comment deleted]

Please correct me if I’m wrong, my understanding is that you are setting the "languageHints" property to Arabic(“ar”) for detecting the English text within the image you provided(478335475.pdf1.pdf10.png). I tried to reproduce your case as described and did obtain less accurate TEXT_DETECTION results than when specifying a "languageHints": ["en"] property.

As described in the Image Context documentation[1], a wrong language hint may prevent the text detection feature of the Cloud Vision API from working appropriately and so produce inaccurate results. And you are right concerning the Google Drive detection mechanism, the Drive Help documentation[2] states that the language is detected automatically. Therefore, in order to obtain similar results to Google Drive(through a Docs file) when using the Vision API to detect text, you would need to leave the "languageHints" blank and let the OCR automatically detect the language.

[1] https://cloud.google.com/vision/reference/rest/v1/images/annotate#imagecontext
[2] https://support.google.com/drive/answer/176692?hl=en

in some image we have two language like attached image.
when i leave "languageHints" blank,it will detect japan language but this image language is fa+en.when i set fa in "languageHints" it will be correct but output is not good than Google docs.
i think language detector of text_detection is too weak in AR+FA and other Asian language.it have good output but when users convert many page they should edit every page.if user want to edit that he/she will type it again and OCR will be useless.
i used Google drive API for convert.drive API convert time is too long ( 15 sec average ) and sometime it show unknown error(like 500 error ( Server Busy Try again) ).
i create a FA+AR+EN OCR with opencv and MLP neural network.it have better output than Google cloud but i can't create many sample for learn in neural network and for this reason i had to use Google cloud ( i worked on machine vision and my OCR yet).
how can i fix text_detection problem and get output like Google Docs?
can Google team fix this problem?

I have tested the last image you provided and did get different text recognition results while using the Google Drive compared to the Vision API TEXT_DETECTION functionality.

I see how this would be a nice feature to be able to detect text with Google Vision API the same way as with Google Drive. You'll be glad to know that I've already submitted a feature request regarding this to our backline team.

While I'm unable to provide you with a timeline regarding when the feature will become available, I want to let you know that such requests are taken very seriously and reviewed regularly for possible implementation on a rolling basis.

Concerning your issue about the latency of Google Drive for converting images to text, you may get faster results by resizing the images as described in this Image sizing link[1]. As an example, I was able to get a much faster conversion time with similar accuracy when I resized the last image you provide to a 1024 x 1499 Pixel size.

[1] https://cloud.google.com/vision/docs/best-practices#image_sizing

[Comment deleted]

hi,
thank you for your attention to resolve this problem.
i hope to google team solve this problem to users of google cloud service.
i use text_detection service in my website but users are unhappy when
see output.
about google drive,when i use google drive API in sometime it will be
very slow ( even small size image ) and sometime show unknown error
and had to do convert again with random wait number.so it may take 30
seconds to convert every page.
i closed my website temporary and waiting to google team solve
this.because with this problem i I lose my credibility.
thank you.

Hi Mojtaba,

I wanted to let you know that I received feedback from the backline team and they accepted the feature request.
Again, note that such requests are reviewed for possible implementation on a rolling basis, as a result there is no ETA on its implementation and any updates will be posted here.

Hi,
I do not know how to thank you.
So i waiting to Google back-line team resolve this.
regards.

Hi,
Is the problem solved?
Changes have been done?How long should I wait?
Thank you ;)