1. Building a Deep Neural Network to play FIFA 18 FIFA 2018 is on. And for those of us who are not privileged enough to be part of the actual teams who play the game, we might as well resort to the FIFA 2018 video game. This tutorial by Chintan Trivedi, published on Medium, guides you through the mechanism of playing football, and how to train a neural network bot to play the game. It walks you through Convolution Neural Network (CNN), Long Short-Term Memory Networks (LSTM), Supervised learning and the code for the project is also available on Github.
  2. Using Deep Q-Learning in FIFA 18 to perfect the art of free-kicks This project is a follow-up to the aforementioned one, built by the same person. It covers the Q-learning model, reinforcement learning and uses the Tensorflow python library for deep learning implementation and pytesseract for Optical Character Recognition (OCR). Football free kicks aren’t too tricky. It only gets tricky when you try to train a bot to do one. This project is both an interesting and educative one. It’s well documented, comes with results and entertaining videos. Definitely worth a try. The code can be found here. Also Read: NVIDIA’s AI robot can learn by simply observing humans
  3. Using Machine Learning to predict soccer scores Time to take a break from playing and switch to statistical analysis. We make predictions all the times. We are always trying to predict scores, including in football, which is actually pretty much unpredictable. But using a machine learning model and existing data from 30000 games, we could try and create a model capable of predicting the outcomes of upcoming games. I can see how this will be useful, especially in this season where everyone is wondering which team will win and which won’t. There is extensive data mining from different sources, and the model has an accuracy of 70%. Not bad.
  4. Python project on RoboCup 2D Soccer Simulation League This is an A.I meant to play a 2-dimensional football game. The code for this project covers many of the basic strategies of the game. There includes kick-off, whistle, offense, defense, player maneuver and a neural net with both supervised training and unsupervised pre-training. Environmental variables are also taken into play; the nearest to the ball, teammates, goal post and enemies, clearness of the path and whether or not the ball is kickable. Soccer.py was a useful agent employed in the backend of this.
  5. Match & Player Analysis in Soccer: Computer Coding and Analytic Possibilities (PDF)  This project is extensive and includes a lot of sporting details. Sifting through the report, one can spot our area of interest. An extract from the report tells us all we need to know: “The purpose of our project is the full and automated analysis of soccer games. Starting from the video recording of a soccer game, the 2D coordinates of the players on the field and the 3D coordinates of the ball during the whole game are extracted by image processing and stored in a database. The interesting actions from the videos can be labeled for easy retrieval and statistics. The coordinates database and the annotation list are then processed and mined for relevant information. [accessed Jun 19, 2018] There you have it. A list of projects that will make this football season more exciting for the A.I enthusiast in you. While the actual players are catching fun all the way in Russia, we too can get some excitement from the comfort of our computers.