DIKU Bits: 3D Reconstruction with Neural Networks

Speaker

Robin Bruneau, PhD student in the Image Analysis, Computational Modelling, and Geometry (IMAGE) section.

Title

3D Reconstruction with Neural Networks

Abstract

3D reconstruction is used for many purposes: special effects, virtual reality, medical scans, etc. However it is also used in archaeology and entomology for purposes as in my PhD thesis on the reconstruction of insects trapped in amber.

In this presentation, we’ll look at the classic methods used to reconstruct objects/scenes from images, and then we’ll take a closer look at the network-based methods that have made their appearance in the 2020s. To finish, we will see general ideas for adapting these methods to take refraction into account when dealing with insects in amber.

Which courses do you teach?

I am only TA in Numerical Optimisation (NO).

Which technology/research/projects/startup are you excited to see the evolution of?

Blender, Mitsuba3, and the research about fully inverse rendering pipelines.