When I run this image containing various vulgar fraction symbols through the Google Vision demo, it recognizes all of the characters correctly except for those symbols. The same is true when I consume the API whether it be with TEXT_DETECTION or DOCUMENT_TEXT_DETECTION. Is there some way I can configure Google Vision to accurately recognize these symbols?
Vision API doesn't have a specific configuration to detect this kind of characters. The OCR feature has been developed to perform text recognition; however, in some cases, the fraction symbols are not identified as characters, making the service to get wrong values or even they don't recognize the symbols themselves.
Based on this, I have run some tests by using this image and I could confirm that the service was able to recognize several fraction symbols correctly. Keep in mind that Vision API use machine learning models that are being trained constantly in order to improve the results quality.
I suggest you take a look on the Supported Images document where you can find some file format and image sizing recommendations that may help you to improve your results accuracy, as well use the Send Feedback button located at the lower left and upper right corners of the Vision API public documentation in case you want to provide your opinion about the product.
