Introduction
March madness started this past week and has college basketball fans buzzing around the USA. For those unfamiliar march madness is a college basketball tournament held every year in the USA where college basketball teams compete to be crowned national champions to conclude that year’s college basketball season. In recent years, however, there has been a major push of data and data analytics in march madness – much like other sports – to decide which play to run and which shot to take. Today we will peel all of that back and see how to even start to get there. First and foremost, how do we identify which teams are on the court to then use to predict which shots to take and when to take them?
Summary of Article
Well in an article written by Priya Dwivedi on the subject she explains how to take game footage and distinguish which players are on which team with data science (the article can be found here). First, she starts by detecting all the people on the court by using a model called Faster RCNN and runs frames of the game through that to get the detections of the people which this model has been trained to do. Then for detecting teams she uses OpenCV to detect the color of the player jerseys and uses that to classify teams. This model firsts masks the image with a color and then references that with the color that you want to detect to detect it. To identify the people shooting a basket she uses the wrist key point and finds players who have that key point above their head and then checks if the ball is above their head and next to them to ensure they are actually shooting and doing something else.
My Take
I really liked this article as it gave me an introduction to data science and deep learning and how that can be used in more fun things like sports. I do however wonder how the model actually detects the humans or finds the key points on the players so I will definitely be reading up more on that. The article was also very well written – organized into multiple paragraphs with headings making the article easy to follow. The content in the article was well explained and further links were provided for people who are interested (me!), which is always helpful. Sports I feel are a really good bridge for data analytics and data science into the mainstream as sports are very interesting to the average person, and when data science and analytics have a role to play in sports especially when the role is telling coaches what plays to pick the average fan will be intrigued and want to learn more about it.
Conclusion
All in all, I really liked the article. It was written, organized, and explained well, and the author provided extra links for people who are interested, which I appreciate. I also feel that data science and analytics in this industry have made a huge impact and due to this people will start learning more about them. Overall it was a really good article and I recommend you read it(the article can be found here).