Project for automating the video game Finding Mii using modern computer vision. You can see the full details on the technical paper here.

This paper details an attempt to automate the Find Mii challenge. We use several techniques like face detection, face recognition and other computer vision algorithms. We attempt to construct algorithms that are capable of: identifying a given avatar in a group of avatars; identifying two look-alike avatars in a group of avatars; identifying an avatar whose behavior differs from that of the rest of the avatars in a group and identifying the fastest moving avatar in a group of avatars. Each of these tasks has three levels of difficulty easy, medium and hard where each of the levels pose unique algorithmic challenges.

This project was done for a graduate computer vision course at Stanford (CS 231A, Fall 2012)