Architect-Ant: Editable Automatic Furnishing of Architectural Floor Plans
Architect-Ant learns professional furnishing from 505 real floor plans and generates editable furniture layouts.
Architect-Ant generates furniture layouts for architectural floor plans using an editable coordinate-based domain-specific language. It is trained on AntPlan, a dataset of 505 real professional plans with dense annotations across 92 object classes and ten residential room categories. The model first learns furnishing patterns by supervised fine-tuning, then is optimized with GRPO using a Layout Rule Score that aggregates geometric and functional constraints. Experiments report low geometric violation rates, high functional completeness, and layouts that remain object-level editable and convertible to 3D.
- AntPlan contains 505 annotated professional floor plans and 92 furniture classes.
- Architect-Ant uses an editable coordinate DSL, supervised fine-tuning, then GRPO.
- A Layout Rule Score enforces geometric and functional constraints.
- Layouts stay object-editable and can be converted into 3D scenes.
From the original article
1 min read at huggingface.co
Furnished floor plans support real-estate visualization, interior design, and architectural workflows, yet automatic furnishing remains challenged by limited real-world data and the need to satisfy interacting geometric and functional constraints.
By Fedor Rodionov, Aleksandar Cvejic, Michael Birsak · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.