HARMONY: Hierarchical Agentic Reasoning for MONocular Image-to-Scene Synthesis
HARMONY reconstructs indoor 3D scenes from one image using hierarchical agentic reasoning and geometry refinement.
HARMONY reconstructs a complete indoor 3D scene from a single monocular image by combining agentic vision-language reasoning with visual geometry models. Starting from an empty floorplan, it calibrates the camera, recovers room layout and an initial placement order, then inserts wall-mounted elements, free-standing furniture, and dependent decorations using depth-first traversal and a feedback loop. After each vision-language placement, point-cloud estimates refine geometry so the render better matches the input. Experiments on synthetic and real images outperform the evaluated reconstruction baselines, and qualitative comparisons with GPT-6 Astra show more faithful object arrangements and scene detail.
- Hierarchical agentic reasoning reconstructs indoor scenes from one image.
- Objects are placed from walls and furniture down to decorations.
- Point-cloud refinement aligns each placement with the reference photo.
- Qualitative results are compared with GPT-6 Astra.
- AI models
- GPT-6 Astra
From the original article
1 min read at huggingface.co
Compositional 3D scene reconstruction has recently been explored from two directions: agentic reasoning that provides semantic understanding of spatial relationships but lacks precise alignment with input images; and visual geometry foundation models that predict dense point maps from input images but the reconstruction quality is limited.
By Shufan Sun, Chen Wang, Enxin Song · 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.