Worlds Unseen: Inferring Hidden Structure with Generative Vision Models. When observing the 3D world, the unseen part can only be surmised from experience. This project aims to infer the hidden struct

Description

Worlds Unseen: Inferring Hidden Structure with Generative Vision Models. When observing the 3D world, the unseen part can only be surmised from experience. This project aims to infer the hidden structure of the 3D world by learning from generative computer vision models trained on large image datasets. This will facilitate the reconstruction of realistic and complete 3D models from images. It will generate new knowledge in computer vision using innovative techniques from differential geometry. Expected outcomes include open-source software for intelligent agents, such as robots, that better understand the 3D world in order to perform more useful tasks. Significant benefits to the manufacturing, aged care and transportation sectors are expected, with applications in robotics, smart homes and autonomous vehicles.. Scheme: Discovery Early Career Researcher Award. Field: 4603 - Computer Vision and Multimedia Computation. Lead: Dr Dylan Campbell