You may have heard recently about the recent "AI Controlled Camera" that mistook a referee's bald head for a ball and began following his head instead of the ball.
1. First, the obvious. It's fascinating that this soccer club has replaced a camera operator with an AI controlled camera. It appears that this club is a lower tier Scottish soccer club and has found a way to televise their games without having to employee and pay a camera operator.
I have not watched any of their games to be able to say how effective this algorithm is at following the run of play, but given that I do watch quite a bit of soccer I will say that the majority of the game is about following the ball. I mean that's where the action is, so I would assume using the ball following algorithm would be the correct one.
To reiterate, what a cool application of AI/ML.
2. Where did it go wrong. Image classification algorithms typically use a neural network that trains on hundreds of images that have been properly labeled and then split into training/testing sets. I would imagine when they were building the classifier, someone labeled those soccer balls, potentially making the assumption that the home soccer balls would all be the same or very close. I would also make the assumption that there was not a bald head in the training-set to be labeled as such. The good news is this should be a relatively easy fix for the team to make.
When you are using your own data to help you become better informed; you need to have confidence in the data. When you do, you can leverage it to do very powerful things. We believe that your real estate data when combined with additional sources can be leveraged to give you superpowers and further embolden your data-driven deal strategy.