Audio-visual object-based production for live events.

点击翻译按钮获取中文摘要

Audio-visual object-based production for live events. Signal processing and machine learning techniques are useful forextracting and tracking the metadata that accompanies media objects in an object-based workflow. Higher-level labelsmay tag the video category or genre giving semantic information about a piece as a whole, where complementary datafrom different modalities tends to enhance usability. In contrast, this project interrogates the more-detailed information thatcan assist in the construction and manipulation of content within an authored audio-visual experience, such as the relativelocation and geometry of actors or other scene elements. Furthermore, the focus will be towards techniques applicable ina live production setting, e.g., for immersive and interactive 3D audio-visual media.