1
$\begingroup$

I am trying to work on a problem to extract the brand names from images of different alcoholic beverages. My dataset consists of images of bars, restaurants, and clubs serving alcohol. Any suggestions on how to achieve this?

Sample Image

For example, I am trying to capture the name Jack Daniels from the above image.

$\endgroup$
  • $\begingroup$ Provide some example pairs of an image and the desired output so we're sure what you mean. $\endgroup$ – Kodiologist Nov 15 '17 at 0:30
  • $\begingroup$ That link gives me a 404 error. Stack Exchange provides a way to upload images, so you could try editing your question and adding images there. $\endgroup$ – Kodiologist Nov 15 '17 at 4:22
  • $\begingroup$ @Kodiologist Thanks for the tip. Hopefully, this helps. I am trying to capture the name Jack Daniels from the above image. $\endgroup$ – Amol Modi Nov 15 '17 at 21:08
  • 2
    $\begingroup$ I don't think I know enough about computer vision to give a good answer, but my guess is that this problem is very difficult and can only be solved with a truly enormous amount of training data, like how Google reads street signs. Existing OCR software won't help because the brand names are stylized and there's competing text in the image. If there are only a few (say, 20) brand names you need to recognize, however, it could be tractable—you could try seeing which of the 20 logos (in standard forms) match to an input image the best. Any other constraints could make a big difference, too. $\endgroup$ – Kodiologist Nov 15 '17 at 21:22

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.