Pose detection and generation with a web camera
– Human poses are detected using the web camera, features such as distance, speed, and confidence values are fed into a pre-trained CVAE network to generate textural images. which are then decoded as skeletal movement data. Simultaneously, generated texture is re-synthesised by converting shader information into an audio buffer, which can be heard using the interactive demo below.
– This online tool is designed to break down and illustrate the core interactive elements of the generative system used in the Temporal Spaces project. This operates using far more simplified mechanisms, leveraging web frameworks and generative modelling, leading to interesting yet unpredictable results.
Pose detection and generation demo https://modina-eu.github.io/Temporal-Spaces-Pose-Visualizer/
Pose detection and generation source code https://github.com/modina-eu/Temporal-Spaces-Pose-Visualizer
Webcam permissions are required for pose detection to work
Source code and documentation for the project’s production software is available in the following repository, including Unity3D, VCV Rack, and vvvv projects: https://github.com/modina-eu/Temporal-Spaces/tree/main