On-Device Sensor-Based Human Activity Recognition Applications
This exhibit demonstrates the shortcomings of current datasets and machine learning models used to detect human activity such as walking, running, and sitting and offers solutions which include a mobile app and a fine tuned machine learning algorithm that uses continual learning to help move a human activity recognition system into real-world, long-term deployment.
Keywords
Deliverable for individual master’s project, Data Analysis, Supervised Machine Learning, Data Collection, Mobile Application, Continual Learning, Edge Computation