In this talk I will present my recent research, especially representing functions on manifolds through neural networks (the equivalent of NeRF but on manifolds) and using quantum annealing to solve problems in computer vision. First, we will look into how the function approximation power of neural networks can be used to define a continuous and differentiable texture representation on 3D shapes, how to optimize this and the theory behind why it works so well. The next part will be a short introduction into quantum annealing and its unique properties, how it can be used to solve correspondence problems, and why combining it with learning approaches makes sense.
17/04/2023