top of page
  • brigitterohner

A novel class of networks for capturing dynamic 3D scenes

Our partners at ETH Zurich have addressed the limitations of multi-layer perceptron (MLP) when modeling 3D data. Marko Mihajlovic, Sergey Prokudin, Marc Pollefeys and Siyu Tang propose incorporating temporal residual layers into neural fields, dubbed ResFields, a novel class of networks specifically designed to effectively represent complex temporal signals.

ResField layers incorporates time-dependent weights into MLPs to effectively represent complex temporal signals.

Capturing dynamic 3D scenes from sparse sensory inputs of a lightweight capture system is an essential part to enable virtual surgery participation and teaching.

Have a look and read the pre-print.


bottom of page