BrickAnything: Geometry-Conditioned Buildable Brick Generation with Structure-Aware Tokenization

The generation of physically constructible brick structures from 3D shapes requires more than simple geometric reconstruction, as it must also satisfy discrete part constraints and structural stability. Current brick generation methods rely on heuristic optimization, which can fail when the target 3D shape does not allow for a feasible structure under predefined constraints, or generate brick sequences without explicitly modeling underlying geometry and assembly relationships. This work presents BrickAnything, a geometry-conditioned autoregressive framework for generating constructible brick structures from various 3D representations. BrickAnything uses point clouds as a unified geometric interface and predicts brick sequences that reconstruct the target shape under assembly constraints. To model structural dependencies between bricks, a structure-aware tokenization is introduced, representing brick structures through local union relationships. This makes sequence generation more consistent with physical construction and reduces invalid intermediate states. This development is significant as it demonstrates how AI can improve brick structure generation, potentially applicable in architecture and engineering. Moreover, the ability to efficiently generate constructible brick structures can be useful in developing solutions for online marketplaces of second-hand goods, such as the open-garage project by dataqbs, aiming to improve the way used objects are bought and sold.

Read the original article on arXiv cs.AI

This summary is an informational synthesis produced by dataqbs.com. All rights to the original content belong to its author and the cited media outlet. We act solely as curators of technology news and claim no authorship.