In activity recognition, the goal is to automatically infer activities or actions of human subjects from observations. Observations can include visual information such as images or videos, acceleration sensors worn by a subject, sensors that detect the interaction with objects, or other indirect measurements from which activities can be reconstructed. State-of-the-art activity recognition approaches are usually based on machine learning, where a model is learned from data to recognize specific activities. In this seminar, students will present and discuss different approaches to carry out activity and action recognition based on machine learning.