Computer vision has become more important in the past few decades having a wide variety of application. Many of the current models available are very domain specific. This project looks at a PhD thesis[8] where the paper’s goal was to explore general computer vision algorithms which can easily adapt to any domain without interference. The algorithm would adjust itself according to the data it is trying to model instead of being so rigid. The proposed model utilized advancements in neural networks and unsurprised learning techniques to track moving objects.