3D Scene Undertsanding
Connecting between the physical and virtual realities, by deriving meaning from topology.
Transforming a physical environment into a virtual one is a challenging task, allowing autonomous agents to traverse and interact with the environment, as well as bridging between users immersed in VR with their surroundings.
This work was focused on decomposing a 3D scene, represented as point cloud or mesh, into meaningful objects, using a combination of algorithms based on differential geometry.
Developed for Resonai, which aims to give machines human-like (or super-human) 3D understanding, with applications that include VR/AR content creation, robotics and more.
I was involved in the research and development of several algorithms used for the decomposition, as well as evaluating the different approaches against an academic benchmark, and deploying the production-ready solution on the cloud.