The objective of this research project is to improve the state of computer vision based localization. The current software allows a user to take a picture of an area and send it to a server, which will match the picture against known data and return location data to the user. There are cases in which similar looking places result in incorrect matches. The methods involved in resolving this include writing software that finds images in a dataset that are easily confused with one another and use that data to make localization guesses more accurate. It may also involve porting Matlab code to C to improve the speed of matching. The anticipated result is having the server return fewer incorrect results.