CVI²

We invite paper submissions to the 4th edition of the SHARP (Solving CAD History and pArameters Recovery from Point clouds and 3D scans) Workshop in conjunction with ICCV 2023 to be held in Paris on October 3rd, 2023. Designed to foster the development of 3D modelling and processing techniques that exploit the geometric, language-based, and parametric representations of 3D scanned data, the SHARP Workshop will solicit new methods for recovering the history and parameters of Computer-Aided Design (CAD) models from 3D scanned data. In addition to scan-to-CAD problems, all topics that relate to and serve the goal of data-driven models for 3D shape parametrization, language-based 3D shape modeling/generation, and 3D shape procedural modeling are of interest. Topics of interest include, but are not limited to:

Prof. Daniel Ritchie

Short Bio: Daniel Ritchie is the Eliot Horowitz Assistant Professor of Computer Science at Brown University, where he co-leads the Brown Visual Computing group. His research sits at the intersection of computer graphics, artificial intelligence, and machine learning: he build intelligent machines that understand the visual world and help people be visually creative. Much of his group’s current work focuses on analyzing, synthesizing, and manipulating 3D scenes and the 3D objects that comprise them.

Abstract: Programs which generate 3D shapes have a variety of applications, from procedural modeling for games and animation to computer-aided design for manufacturing. The ability to infer a generative program for a given 3D shape would be valuable, enabling reverse-engineering and procedural editing of 3D objects found in the wild. Nowadays, machine learning approaches are the dominant paradigm for solving such inference problems. Unfortunately, training data in the form of (3D shape, generative program) pairs rarely exists, making supervised learning challenging or impossible. In this talk, I’ll discuss unsupervised approaches for solving the 3D shape program inference problem, i.e. methods that assume no access to such paired training data.

Prof. Lourdes Agapito

Short Bio: Prof. Lourdes Agapito is a Professor of 3D Vision at the Department of Computer Science, University College London (UCL). Her research in Computer Vision has consistently focused on the inference of 3D information from single images or videos acquired from a moving camera. She serves regularly as Area Chair for the top Computer Vision conferences (CVPR, ICCV, ECCV). Her perspective on the next steps in 3D data parsing and representation learning will motivate interesting discussions amongst the SHARP community.

Abstract: TBA