Use Cases for Spotify API in relation to Machine Learning

[njd57]

My Question or Issue

Hello all, quick question about use of the Web API for developers.

For background, my roommate and I are Senior Engineers in college, working on our capstone project. Our idea involves using sensors & a camera to collect data about a room, and use a machine learning algorithm to decide what song(s) to play. This project is not monetized in any way, not for profit, and otherwise doesn't go against any rules laid out in the developer agreement.

From our understanding, feeding data from the API into a machine learning model is strictly not allowed. To remedy this, we have decided to have the machine learning aspect of our project be wholly removed from the API calls. The idea is to have the data parameters from the computer vision portion be fed into the machine learning algorithm, and then use some other arbitrary metric like genre to be the output. From these arbitrary outputs, we will then query the API for recommendations. 

Since the Web API data is in no way directly involved in the training of the model (neither input nor output), would this use case be acceptable? Additionally, if this is allowed, what are the limits of using data metrics that are similar to data returned in the song info - i.e. tempo or other defining traits of the songs? The Spotify Web API seems to be the only real option to bring this project to life, and if this isn't allowed or needs to be altered it would be better for us to know sooner rather than later so we can rework the requirements. Thanks in advance for any help/advice.