If no measures are in place to block a name/size combination from multiple guesses (don't see why they would implement this) then I would just setup 4 bots. Each to answer with #1, #2, #3, or #4 - those hashtags can be OCRed pretty easy out of the box.
I'm guaranteed that multiple entries will not get admitted because there is only one right answer. So unless you had more than 1 variation of the image base question being asked - it wouldn't be hard to beat. Just need 4 bots for each name/size entry.