Book and Table of Contents

I found it works best when you watch the video and use the book as a reference.

If you are an experienced data or sport scientist or a hobbyist, looking to understand AI better, this book should give you plenty of inspiration and practical examples.

-- Phil Cheetham, PhD, US Olympic team sport scientist and twice Olympian


Table of Contents

Machine Learning in Sports 101 (Supervised, unsupervised, reinforcement learning – Logic and machine learning – Tools – Neural networks – Deep vision – Classification – Detection – Semantic segmentation – Sensors – Reinforcement learning)

Physics of Sports (Mechanics – Kinetics – Laws of motion – Inertia – Kinematics – Projectile motion – Using neural networks to predict a projectile range – Angular motion – Conservation laws – Energy, work and power – Physics and deep learning)

Sports Scientist Toolbox (Data science tools – Python – Virtual environments – Packaging – Anaconda – Pip – Jupyter notebooks – Numpy – Pandas – Visualizations – Matplotlib – SciPy – Scikit-image – OpenCV – PyTorch – Keras – Tensorflow – OpenAI Gym – Pybullet)

Neural Networks (Neurons – Activation – Perceptron – Training a perceptron in Python – Multi-layer networks – Backpropagation)

Sensors (Deep Vision – Edge devices – Inertial movement sensors IMUs – Attitude and heading reference systems AHRS – Inertial and navigation systems GNSS – Range Imaging Sensors LIDAR – Pressure sensors – EMG sensors – Heart rate sensors )

Deep Computer Vision (Neuroscience and deep learning – Computer vision – Visual datasets – Model zoo – Applying models – Classification – Classifying sport activity type – Detection – Segmentation –Semantic segmentation – Human body keypoint detection)

2D Pose Estimation (Methods – Neural networks – Datasets – Tools – Body pose estimation – Detecting athlete stance – Activity recognition – Detecting skill level – Multi-person pose estimation – Dealing with loss and occlusion)
3D Pose Estimation (Cameras and 3D – Camera Matrix – 3D Reconstruction – Using a single camera – Multi-view depth reconstruction – 3D reconstruction with sensors – Motion capture – 3D Datasets – 3D Machine learning methods – Sparse and dense reconstruction)

Video Action Recognition (Video Data – Datasets – Models – Video classification – Action recognition – Loading videos for classifier training – Visualizing datasets – Video normalization – Training video recognition model)

Reinforcement Learning (Tools – Applying reinforcement learning in sports – Action and observation spaces – Visualizing sample motion – Model zoo – Models – Reinforcement learning in gymnastics – Pendulum model – Humanoid models – Joints and action spaces – Human motion capture – Mocap – Reinforcement learning in humanoids)

Machine Learning in the Cloud (Containers – Docker – Notebooks in the cloud – Data storage and datasets in the cloud – Loading and accessing datasets – Labeling data in the cloud – Training classification model – Preparing for training – Running experiments – Model management)

Automating and Consuming Machine Learning (CI/CD – MLOps – Managing models – Creating a scoring script – Defining an environment – Deploying models – Consuming models – APIs – Machine learning pipelines)