Enhancing high-resolution 360 streaming with view prediction

What this post added

Introduced a gravitational view-prediction model using physics and heatmaps to improve resolution on VR devices by up to 39 percent. Developed an AI model to generate saliency maps for videos, enabling prediction even without heatmap data. Introduced content-dependent streaming, a new encoding technique that improves 360 video resolution on non-VR devices by up to 51 percent, even in low bandwidth, by transforming frames based on saliency data into a single stream.