Videos are recorded simultaneously with sensor data (IMU) to provide ground truth labels that are required for training an AI, which can automatically recognize animal activities from IMU data. Traditionally, the videos are only used to annotate the IMU data with labels (supervised learning). Using the raw video data to enhance the AI classifier is a tremendous opportunity.
The student will investigate how to leverage additional data during training, which is not available during deployment. For example, train an AI classifier with video and IMU data during training and use only IMU data during testing. Real-world data comprising IMU data and videos from animals such as horses, goats, and sheep can be used for this assignment.
20% Theory, 60% Simulations, 20%Writing
Jacob Kamminga, email@example.com, room ZI5011