Strokes2Deform: Physics-informed learning of deformation fields on 3D stroke clouds

Computers & Graphics 2026 Open access

Read full text
Authors: Shervin Rasoulzadeh, Raman Suliman, Arvin Rasoulzadeh, Iva Kovačić, Michael Wimmer
DOI: 10.1016/j.cag.2026.104629
Cited by: 1
References: 27

Abstract

• A learning-based model that enables deformation-aware 3D sketching without requiring reconstruction-and-simulation. • A synthetic dataset of 40K sketch-deformation pairs spanning 3D sketches of architectural thin-shell structures and their corresponding deformation fields derived from Finite Element Analysis (FEA). • A dual-head physics-informed neural network that decouples deformation field estimation into predicting unit-length displacement vectors and scalar displacement magnitudes. • Physics-guided regularization terms incorporating stretching and bending as the two modes of deformation.

Fields of Study & Subjects

Model Reduction and Neural Networks Machine Learning in Materials Science 3D Shape Modeling and Analysis

Publication details

Journal / source
Computers & Graphics
Publisher
Elsevier BV
Volume
138
Pages
104629–104629
Publication date
26 May 2026
Language
EN