Massachusetts Institute of Technology

Vendredi 19 janvier, 11:00-12:00, Salle 3195, Pavillon André Aisenstadt

Cette présentation sera en anglais.

Geometric modeling is a core component of tools for manufacturing,
design, entertainment, medical imaging, and computer vision. Despite the
demand for 3D models in these disciplines, however, creating digital
geometry remains costly, time-consuming, and specialized. The core
problem of modern modeling software is that modeling is presented as a
sequence of operations with tools. For instance, an artist looking to
design a humanoid character must build up its shape using a sequence of
operations like curve creation, extrusion, and smoothing. For
comparison, when one human wants to communicate an idea of a particular
shape to another, such tool-oriented explanations would be excessively
time-consuming: there are much easier ways.

In this talk, we will explore an alternative modeling paradigm based on
natural---yet often ambiguous---input, such as hand-drawn sketches.
Within this paradigm, we will outline future directions for 2D and 3D
modeling systems, as well as present projects paving the way toward
intuitive software for modeling shapes.

Biographie :

Mikhail Bessmeltsev is a postdoctoral associate at Massachusetts
Institute of Technology (MIT), working with Prof. Justin Solomon.
Mikhail completed his PhD in Computer Science at the University of
British Columbia (UBC) under supervision of Prof. Alla Sheffer. Before
that, he led a research and development group within the Novosibirsk
State University (NSU) Intel High-Performance laboratory. This group
later formed the core team of a laser-scan processing startup, where
Mikhail served as CTO. Mikhail's scientific interests span topics in
computer graphics, vision, and machine learning, including geometry
processing, animation, and modeling.

Venez nombreux !