ARAPReg: An As-Rigid-As Possible Regularization Loss for Learning Deformable Shape Generators

TL;DR

This paper presents ARAPReg, an unsupervised regularization loss that enforces local rigidity preservation in deformable shape generation, improving the quality of generated shapes across various categories.

Contribution

It introduces a novel ARAP-based loss derived from spectral decomposition, enabling easy integration into standard generative models for better shape preservation.

Findings

Outperforms existing shape generation methods on benchmark datasets

Easily integrates with models like VAE and auto-decoder

Effective across diverse shape categories such as human, animal, and bone

Abstract

This paper introduces an unsupervised loss for training parametric deformation shape generators. The key idea is to enforce the preservation of local rigidity among the generated shapes. Our approach builds on an approximation of the as-rigid-as possible (or ARAP) deformation energy. We show how to develop the unsupervised loss via a spectral decomposition of the Hessian of the ARAP energy. Our loss nicely decouples pose and shape variations through a robust norm. The loss admits simple closed-form expressions. It is easy to train and can be plugged into any standard generation models, e.g., variational auto-encoder (VAE) and auto-decoder (AD). Experimental results show that our approach outperforms existing shape generation approaches considerably on public benchmark datasets of various shape categories such as human, animal and bone.