The initial idea was to be able to search Google images for symbols such as Chinese characters. It turns out this search fails miserably. Seemingly insignificant things like stroke thickness and color (vs. black & white) can significantly alter the results. For instance, if you draw a
π you may or may not get to the pi symbol - most of the times it fails. So this quickly turned into a game to see which doodles Google was able to recognize correctly. This was pretty fun, and it works pretty well if you draw e.g. emojis or fruit. Of course your use case may be very different.
General user flow
The sequence diagram below explains the general user interaction flow.