A 4D Foundation for Deformable Objects

Brief description of the context of the research including potential impact
This is scene reconstruction in the field of computer vision. Augmenting techniques that can recreate the 3D world from images within the scene to include the capability to represent objects which change through time such as through articulation or deformation. This concerns animals, people, and human created objects, such as shopping bags or construction bags. This work could fit into a robotic software stack to locate deformable objects in their view. Other applications include avatars for augmented reality calls.
Aims and Objectives
Enhance methodology for reconstructing and representing deformable objects using neural network based machine learning models.
Novelty of the research methodology
We will explore novel methods (architecture - optimisation, design - and data) for representing deformable and articulated objects. The models will be newly applied to the area. We will use synthetic data and transformers to create generalisable reconstruction models.
Alignment to EPSRC's strategies and research areas (which EPSRC research area the project relates to) Further information on the areas can be found on http://www.epsrc.ac.uk/research/ourportfolio/researchareas/
Graphics and visualisation
Artificial intelligence technologies
Any companies or collaborators involved
n/a