Research map / 3D vision / 3D generation
ShellMaker: Language-Guided Exterior Completion under Structural Constraints
Ruiqi Xu, Daniel Aliaga
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use joint wall-roof texture retrieval instead of independent retrieval (ablation) | ShellMaker · Exterior completion | CLIP-Sim: 0.285 -> 0.315 | Improved |
| use joint wall-roof texture retrieval (ablation) | ShellMaker · Exterior completion | UNI3D: 0.357 -> 0.424 | Improved |
| add LLM part-aware prompt refinement (ablation) | ShellMaker · Exterior completion | CLIP-Sim: 0.293 -> 0.315 | Improved |
| add LLM part-aware prompt refinement (ablation) | ShellMaker · Exterior completion | UNI3D: 0.364 -> 0.424 | Improved |
| add LLM prompt refinement (ablation) | ShellMaker · Exterior completion | Opening IoU: 0.922 -> 0.883 | Hurt |
| use both joint texture and LLM prompt together (ablation) | ShellMaker · Exterior completion | UNI3D: 0.305 -> 0.424 | Improved |
| remove joint texture or LLM prompt (ablation) | ShellMaker · Exterior completion | Footprint IoU: 0.992 -> 0.992 | No clear effect |
| hard scaffold constraints with geometric assembly instead of unconstrained Trellis-2 | ShellMaker vs Trellis-2 · Exterior completion | Footprint IoU: 0.492 -> 0.992 | Improved |
| compare ShellMaker with Trellis-2 on CLIP-Sim | ShellMaker vs Trellis-2 · Exterior completion | CLIP-Sim: 0.318 vs 0.315 | No clear effect |
| apply to CAD scaffolds instead of indoor-generator scaffolds | ShellMaker · Scaffold source | Footprint IoU: 0.992 -> 0.889 | Hurt |
| apply to CityGML scaffolds instead of indoor-generator scaffolds | ShellMaker · Scaffold source | Footprint IoU: 0.992 -> 0.991 | No clear effect |
Closest papers by meaning
- Global Graph-Validated Optimization for VLM-based 3D Indoor Scene Generation (ECCV 2026 · similarity 0.87)
- M3DLayout: A Multi-Source Dataset of 3D Indoor Layouts and Structured Descriptions for 3D Generation (CVPR 2026 · similarity 0.87)
- Direct Numerical Layout Generation for 3D Indoor Scene Synthesis via Spatial Reasoning (NeurIPS 2025 · similarity 0.86)
- OptiScene: LLM-driven Indoor Scene Layout Generation via Scaled Human-aligned Data Synthesis and Multi-Stage Preference Optimization (NeurIPS 2025 · similarity 0.85)
- Tokenization Allows Multimodal Large Language Models to Understand, Generate and Edit Architectural Floor Plans (CVPR 2026 · similarity 0.85)
- HouseLayout3D: A Benchmark and Training-free Baseline for 3D Layout Estimation in the Wild (NeurIPS 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
