05.09.2023We kindly invite you to participate in the seminar Neurosymbolic Models for Materials and 3D Shapes by Dr. Paul Guerrero, Research Scientist at Adobe Research in London.
The seminar will be held on September 12, 2023, at 14:30 at the IDEAS NCBR conference room as well as online.
Abstract:
Recently neural networks have become popular for graphical content creation, including images, materials, and 3D scenes. While it is relatively straight-forward to use these neural models to get highly realistic outputs (given enough training data), they lack controllability and interpretability: parameters of these models are hard to interpret, it is difficult to reliably control their outputs, and it is hard to ensure that outputs are free of undesirable artifacts. Due to these downsides, adoption in the industry has been slow, and procedural models are still preferred over neural models in many cases. Procedural models are highly controllable and interpretable, but it is hard to author a procedural model that produces realistic output. In this talk, I am going to give an overview of my recent work in Neurosymbolic models, which are models that aim to combine the strengths of neural models and procedural models. Specifically, I am going to present neurosymbolic models for SVBRDF-based materials, and for 3D shapes represented by sets of cuboids.
Bio:
Dr. Paul Guerrero is Research Scientist at Adobe Research in London, working on the analysis of shapes and irregular structures, such as graphs, meshes, or vector graphics, by combining methods from machine learning, optimization, and computational geometry. Previously, he was a post-doctoral researcher at the Smart Geometry Processing Group, UCL. He completed his PhD at the Insitute for Computer Graphics and Algorithms, Vienna University of Technology, and at the Visual Computing Center in KAUST.
If you want to attend in person, write us an email at registration@ideas-ncbr.pl
You can also join the seminar online via Zoom:
Link: https://us06web.zoom.us/j/4703485582?pwd=Mml0anpHODVQTWZzTWpUc1huWkt5Zz09
Meeting ID: 470 348 5582
Passcode: 727039
