Update

Meeting notes.

Iso-Points: Optimizing Neural Implicit Surfaces with Hybrid Representations

Introduce an efficient way to recover an explicit representation of an implicit surface in the form of 3D points on the surface to supervise training of deep neural networks that fit a neural implicit function to some data. They show that this helps reduce training time, be more robust to noise and yield result with better topology than without the iso-points supervision.

Zippables

Fabricating 3D objects with fabric that assemble via a zipper. They propose a method to generate a curved path for the zipper on a target surface.

Project Update

Making PEARL segmentation approach more rigorous, improve initialization and add possibility to propose new models generated by splitting subsets into 2. Reflexion on what distance approximation to use for fidelity energy.

Project Math Equations

A place to put all equations in latex format, so that they can be easily reused in the future.

Project Update

Progress on using implicit polynomial surface models

Interactive Liquid Splash Modeling by User Sketches

Using a cGAN to reproduce the effect of a liquid splash, based on some input 3D strokes that indicate fluid flow trajectories and velocity.

3L Fitting

A method to fit implicit polynomial surfaces to data points. They improve stability over other methods by introducing additional constraints in a kind of skirt around the data points, to force the implicit surface to be well behaved in the vicinity of the points.