---
title: "Mira-Scene Rebuilds Editable 3D Scenes From a Single Photo"
subtitle: "A new research paper proposes Canonical Coordinate Maps, a dense pixel-aligned representation that reconstructs full 3D scenes from a single image."
url: "https://alphasignal.ai/news/mira-scene-rebuilds-editable-3d-scenes-from-a-single-photo"
author: "AlphaSignal Newsroom"
author_url: "https://alphasignal.ai/editorial-team"
company: "AlphaSignal"
company_url: "https://alphasignal.ai/company/alphasignal"
published: "2026-09-20T18:36:29.000Z"
updated: "2026-09-25T00:29:48.859Z"
type: "Paper"
topics:
  - "Image"
  - "Video"
subtopics:
  - "Neural Rendering"
  - "Vision Language"
  - "Object Detection"
upvotes: 163
source_url: "https://arxiv.org/abs/2609.23796"
comments: 0
truncated: true
---

# Mira-Scene Rebuilds Editable 3D Scenes From a Single Photo

> A new research paper proposes Canonical Coordinate Maps, a dense pixel-aligned representation that reconstructs full 3D scenes from a single image.

# Mira-Scene reconstructs editable 3D scenes from one image

Researchers led by Yang-Tian Sun describe Mira-Scene, a framework that reconstructs an editable 3D scene from a single image by generating separate object meshes and estimating each object’s position, orientation, and scale. In their [paper](https://arxiv.org/abs/2609.23796), the authors report relative gains of 39.8% in 3D intersection over union and 16.5% in 2D intersection over union against SAM3D. The framework predicts dense pixel correspondences, then recovers object transforms through geometric alignment.

## Why scene layout remains difficult

Single-image generators can produce detailed 3D objects, but assembling those assets into a coherent room, street, or diorama requires consistent depth, orientation, scale, and contact between objects. Small placement errors accumulate quickly: a plausible chair model still fails as part of a scene when it floats above the floor, intersects a table, or has the wrong scale.

| Approach                 | Strength                             | Tradeoff                                          |
| ------------------------ | ------------------------------------ | ------------------------------------------------- |
| Holistic generation      | Models the full scene jointly        | Often reduces object-level detail and editability |
| Compositional generation | Preserves separate, editable objects | Must infer a reliable transform for every object  |

Training data compounds the problem. Object-level 3D datasets provide abundant geometry, while scene datasets with labeled per-object poses remain limited. Directly regressing a small set of pose values from scarce scene supervision can generalize poorly across object categories and environments. Mira-Scene converts layout estimation into a dense prediction task that can use object-level 3D data without requiring scene-level pose annotations.

_Preview only. Full text requires an AlphaSignal Pro account._

## Sources

- [Mira-Scene: Pixel-Aligned Layouts for Generative 3D Scene (arXiv)](https://arxiv.org/abs/2609.23796)
- [Mira-Scene GitHub Repository](https://github.com/VAST-AI-Research/Mira-Scene)
- [Mira-Scene CCM Diffusion Checkpoint on Hugging Face](https://huggingface.co/Yang-Tian/Mira-Scene)
- [Mira-Scene HTML paper](https://arxiv.org/html/2609.23796)