[Paper Review] The Wreath Process: A totally generative model of geometric shape based on nested symmetries

Diana Borsa, Thore Graepel+1 more|arXiv (Cornell University)|Jun 9, 2015

3D Shape Modeling and Analysis12 references3 citations

TL;DR

This paper introduces the stochastic wreath process, a fully generative probabilistic model that represents geometric shapes through nested symmetries using group-theoretic wreath products. By aligning noise processes with hierarchical transformation groups, it enables Bayesian inference of shape generative histories via reversible jump MCMC and Approximate Bayesian Computation, successfully recovering complex symmetric structures from noisy sketches and model-generated data.

ABSTRACT

We consider the problem of modelling noisy but highly symmetric shapes that can be viewed as hierarchies of whole-part relationships in which higher level objects are composed of transformed collections of lower level objects. To this end, we propose the stochastic wreath process, a fully generative probabilistic model of drawings. Following Leyton's "Generative Theory of Shape", we represent shapes as sequences of transformation groups composed through a wreath product. This representation emphasizes the maximization of transfer --- the idea that the most compact and meaningful representation of a given shape is achieved by maximizing the re-use of existing building blocks or parts. The proposed stochastic wreath process extends Leyton's theory by defining a probability distribution over geometric shapes in terms of noise processes that are aligned with the generative group structure of the shape. We propose an inference scheme for recovering the generative history of given images in terms of the wreath process using reversible jump Markov chain Monte Carlo methods and Approximate Bayesian Computation. In the context of sketching we demonstrate the feasibility and limitations of this approach on model-generated and real data.

Motivation & Objective

• To develop a fully generative model of geometric shapes based on hierarchical part-whole relationships and nested symmetries.
• To extend Leyton’s generative theory of shape by introducing stochasticity to handle noisy, real-world sketches.
• To enable inference of the generative history of a shape from a single observed image using probabilistic methods.
• To provide a principled framework for shape understanding, editing, and structure recovery through maximization of transfer and reusability of substructures.
• To demonstrate the feasibility of the model on both synthetic and real geometric sketches, including architectural floor plans.

Proposed method

• The model uses wreath products of transformation groups to represent hierarchical shape generation, encoding part-whole relationships through group composition.
• Noise is modeled as independent processes aligned with each level of the wreath product hierarchy, ensuring generative coherence.
• A stochastic rendering pipeline generates pixel-level images from the wreath process, simulating realistic sketch imperfections.
• Reversible jump Markov chain Monte Carlo (RJ-MCMC) is employed to explore model space and infer the most probable generative history.
• Approximate Bayesian Computation (ABC) is used to approximate the likelihood when exact computation is intractable, enabling inference on real sketch data.
• The inference scheme favors simpler explanations with higher transfer, promoting compact and meaningful shape representations.

Experimental results

Research questions

• RQ1 Can a probabilistic generative model based on nested symmetries recover the hierarchical structure of noisy geometric sketches?
• RQ2 How well can the stochastic wreath process infer the generative history of a shape from a single observed image?
• RQ3 To what extent does the model preserve structural regularity while accounting for real-world sketch noise and variability?
• RQ4 Can the model generalize to complex real-world shapes such as architectural floor plans?
• RQ5 How does the principle of maximization of transfer improve shape representation and inference robustness?

Key findings

• The stochastic wreath process successfully recovers hierarchical generative histories for both model-generated and real hand-drawn geometric sketches, demonstrating feasibility on symmetric structures.
• Inference via RJ-MCMC and ABC favors simpler, more transfer-maximizing explanations, even when input shapes contain unexplained pixels or irregularities.
• The model accurately infers regular structures such as regular polygons and grid-like arrangements, especially when predefined control groups are used in the proposal mechanism.
• For architectural floor plans like the Dome of the Rock and Villa La Rotonda, the model identifies underlying symmetric patterns consistent with known structural layouts.
• The approach enables intelligent shape manipulation by allowing editing of substructures that are reused across the shape, preserving global symmetry.
• The model's coordinate system based on wreath product structure provides a natural, interpretable way to describe points within a shape, enabling meaningful spatial reasoning.