Exploring Motion Tracking in Elixir with Bumblebee and LiveView

Katelynn Burns presents a deep dive into using Elixir and LiveView for motion tracking, exploring their potential for creating dynamic, interactive applications. She begins with the role of body movement and its complexities that fascinate her, especially in the context of AI challenges in identifying detailed body structures. Katelynn discusses why Bumblebee and LiveView are apt choices for the task, with Bumblebee providing approachable machine learning tooling via pre-trained models and LiveView's stateful nature being ideal for updating UI elements in real-time without unnecessary overhead. The keynote progresses through the conceptualization and step-by-step development of an Elixir application capable of motion tracking. This application hinges on Bumblebee's neural network capability, NX matrices and operations, and LiveView's open socket design. Despite some challenges with accurate motion prediction using Bumblebee's classification models, Katelynn demonstrates significant progress in the application that leverages contour detection and video manipulation libraries like E-Vision that integrates with OpenCV. By the end of the talk, she encourages the audience to build upon her work and experiment with Elixir and AI to innovate further.