Exercise Classification with Machine Learning (Part I)

The first post will focus on a more algorithmic approach using k-Nearest Neighbors to classify an unknown video, and in the second post, we'll look at an exclusively machine learning (ML) approach. Code for everything we're going to cover can be found on this GitHub repository. The algorithmic approach (Part I) is written in Swift and is available as a CocoaPod. The ML approach (Part II) is written in Python/TensorFlow and can be found as part of the GitHub repository. We want to build a system which takes as input a video of a person performing an exercise and outputs a class label which describes the video.