Research map / 3D vision / 3D generation
Interact3D: Compositional 3D Generation of Interactive Objects
Hui Shan, Keyang Luo, Ming Li, Sizhe Zheng, Yanwei Fu, Zhen Chen, Xiangru Huang
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 |
|---|---|---|---|
| add agentic VLM refinement loop (ablation) | Interact3D · Two-part composition (150 cases) | VLM rating: 8.18 -> 8.57 | Improved |
| add agentic VLM refinement loop (ablation) | Interact3D · Two-part composition (150 cases) | R_volume (x1e-3): 4.9380 -> 3.8762 | Improved |
| add SDF collision term on top of GeoTransformer+ICP (ablation) | Interact3D · Two-part composition (150 cases) | R_volume (x1e-3): 6.0039 -> 4.9380 | Improved |
| use GeoTransformer global init instead of RANSAC (ablation) | registration · Two-part composition (150 cases) | Human rating: 6.40 -> 7.03 | Improved |
| add global initialization (RANSAC) before ICP (ablation) | registration · Two-part composition (150 cases) | R_volume (x1e-3): 9.7102 -> 6.7536 | Improved |
| ICP only versus RANSAC only (ablation) | registration · Two-part composition (150 cases) | R_volume vs R_surface: R_surface 2.2287 -> 1.2914; R_volume 7.9280 -> 9.7102 | Mixed |
| generate-then-compose with registration instead of data-driven baselines (Jigsaw, 2BY2, MIDI) | Interact3D · Two-part composition (150 cases) | Human rating: 5.27 -> 8.33 | Improved |
| Interact3D vs MIDI on penetration | Interact3D · Two-part composition (150 cases) | R_surface (x1e-3): 22.185 -> 0.6939 | Improved |
| Interact3D vs PartField+RANSAC on semantic alignment | Interact3D · Two-part composition (150 cases) | Text CLIP: 0.3094 -> 0.3314 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- MIDI — MIDI: Multi-Instance Diffusion for Single Image to 3D Scene Generation (CVPR 2025) · better on 6 of 6 shared results
- Jigsaw — Jigsaw: Learning to Assemble Multiple Fractured Objects (NeurIPS 2023) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Interactive3D: Create What You Want by Interactive 3D Generation (CVPR 2024 · similarity 0.93)
- WorldAct: Activating Monolithic 3D Worlds into Interactive-Ready Object-Centric Scenes (NeurIPS 2026 · similarity 0.88)
- WonderWorld: Interactive 3D Scene Generation from a Single Image (CVPR 2025 · similarity 0.88)
- Layout-your-3D: Controllable and Precise 3D Generation with 2D Blueprint (ICLR 2025 · similarity 0.88)
- CompGS: Unleashing 2D Compositionality for Compositional Text-to-3D via Dynamically Optimizing 3D Gaussians (CVPR 2025 · similarity 0.88)
- DecompDreamer: A Composition-Aware Curriculum for Structured 3D Asset Generation (NeurIPS 2026 · similarity 0.87)
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.
