According to Cisco, 82% of internet traffic will be dedicated to video [1]. As a result, media are adapting automatic video analysis tools to detect key content, prevent manipulated content, and enforce copyright regulations. This requires the development of video classification, indexing, and retrieval techniques [2]. Despite massive progress in the democratization of video creation, creating a professional looking, cinematic video is very time consuming. Arguably, the most time consuming part of video creation is video editing, which involves selecting the most valuable footage. We can save video editors time by automatically removing unwanted footage; this in turn will allow the video editor to spend less time finding the most valuable footage. We define unwanted parts as technically defective, specified as shaky, blurry, over exposed, under exposed and occluded. Given these classifications can be complex, the scope of this project will be focused on detecting shaky footage.