Is there any reason this can’t be done? Flightaware has a huge pool of photos as well as pilots on there phones. Can we make a program to auto tag tail numbers for flight aware systems.
Google has a text recognition API. Basically, you send the API a picture, and it returns the recognized text, if it finds any. Of course, text recognition is fraught with difficulties, and the hit rate is likely to be pretty low, especially for photos that are not likely to look like sheets of paper.
For an idea of what it will likely recognize, fire up the google translate app, which can translate text found in real time. Obviously, it can do this only if it can recognize the text first.
P.S. I just downloaded an app called Scanner & Translator and tried it out on a few images here on FA. Of the image that I could discern a tail number visually, it recognized about 50% accurately. It would probably do better if you fed it cropped images, cropped around the tail number, but that would seem to defeat the purpose.
One potential gotcha is that it will recognize text not on a plane. For example, one of the images I tested had wording on a building behind the plane. That text was recognized. Also, other lettering not the tail number on the plane will be picked up, too. Recognizing just the tail number from only planes in images is actually a very tricky challenge.
Thats true, I haven’t thought of all the other lettering. Isn’t there more or less a guideline for the shape of t number lettering? Something you could train the api to look for.
It maybe more practical to crowdsource this, and allow members to submit corrections or additions such as Reg, type and location for untagged photos.,
Agreed. How can we make this happen?