Long Live Photogrammetry and SLAM: How Edge Computing and Cheap Hardware Enable Real-Time, Real-World 3D Modeling
Many amazing advances in modeling and visualization have been made possible by the cloud computing renaissance of the past decade. However, even with continual advances in broadband and cellular data speed and availability, cloud access remains a significant impediment in rural, remote, or military environments. But just as cloud has been advancing, GPU technology has opened the door for incredible possibilities on the edge–far from server farms and office buildings. Large-scale data processing is increasingly becoming a reality on local, offline computing hardware, before the cloud is even required.
This presentation will explore just some of the possibilities enabled by recent advances in this powerful but less explored arena of computing power. It will provide a starting point for understanding how GPU technology works, and how it relates to cloud computing, and where each approach might be more appropriate in the world of 3D modeling and advanced visualization. It will discuss how this technology relates to deep learning and artificial intelligence. And it will explore one concrete application of edge computing to advanced visualization and 3D modeling, namely, real-time 3D reconstruction, an application made possible by GPUs and relatively inexpensive commercial drone hardware.
Ultimately, we collect data to make decisions, and yet it is still challenging to handle what is an increasingly massive amount of sensor imagery. This session aims to move beyond hype and provide insight into how edge computing can help address real-world data collection, transfer, processing, and modeling challenges, and how it can be a useful and complementary approach to more popular cloud based approaches.