Thursday, August 25, 2016

Typeface Detection App for Font Finding

Project Status: App Idea

Problem

Sometimes we see the use of an interesting or unique typeface in print or in graphics and we would like to know which font that is exactly. This way we can download it or purchase it for our own use.

Workaround

The current way of tracking down a font is by searching through many fonts online hoping to randomly come across a matching or similar one. Another way is to try to track down the person who designed the graphics or print and attempt to ask this person or team directly.

Solution

I would like to use computer vision, such as OpenCV, and design an app that will not only recognize that there is text in an image/video but will discover the name of the font type used and link to it.

The app will likely be used over and over for some very common fonts and a small database can be made to handle this. By nature though the app will be used in cases of very obscure typefaces and possibly even ones that aren't fonts but the one-off designs used for specific applications.

When a font can not be determined it can be logged. The app user can choose to be notified if the font is labeled at sometime in the future.

Actually, if you have ever used the Shazam mobile app it would work in much the same way as this does but instead of music based data it would focus on fonts.

I would also like it to be a useful tool to those people that design fonts. This app could help them to sell and distribute fonts they own or to popularize ones they give away for free. If done correctly it could also help provide them with user interest feedback, in ways that may help them in designing more fonts. For example; if they see the unlabeled fonts users are searching for they may be able to develop similar fonts that those users may also like.


The tagging process, calculating the probabilities of type matches and even finding similar font types could each largely be done or helped by machine learning. The algorithms and bots could carry the largest parts and the most tedious portions of the workload while their human counterparts could guide them and help to improve the results they generate.

There are several easy to find and scrape font repositories on the web but it would eventually be interesting and useful to write bots to hunt down and include the more obscure ones out there.