Use any corner detector and try to find an convex with some threshold value. Threshold value will be unique for any object. Then based on the threshold value try to find some geometrical properties of the convex. Then you decide?
sounds interesting, as you know there are lot of BRAIN initiatives and a lot of work in robots. Microsoft just shut down a teenage AI bot for becoming a dictator, so be careful. http://www.bing.com/news/search?q=Microsoft+Shuts+Down+AI+Bot&qpvt=micorsoft+shtus+down+AI+bot&FORM=EWRE
Yes, the twitter bot that spouted nazi sympathy is an example of how not to design a chatbot. Problem is, how do you design a bot that has judgement. What we need is a bot design with simulated emotions that are close enough to human to encourage it NOT to bow to social pressure.
Actually, it might have utility for detecting trolls, since it was so easily led down the garden path. Just remove the output capability, and use it to detect over-motivated trolls. If it gets infected with a meme that is obnoxious, trace back the meme to it's source and block tthe base user that introduced or supported that meme. Also remove the offending tweets. I see a role as assistant moderator for it.