---
title: Sat3DGen
emoji: 🛰️
colorFrom: blue
colorTo: green
sdk: gradio
sdk_version: "5.29.0"
python_version: "3.10"
app_file: app.py
pinned: false
license: mit
models:
  - qian43/Sat3DGen
suggested_hardware: t4-small
---

## Sat3DGen – Street-Level 3D Scene Generation from Satellite Images

**[ICLR 2026]** Generate street-level 3D scenes from a single satellite image.

- 📄 [Paper (OpenReview)](https://openreview.net/forum?id=E7JzkZCofa)
- 🌐 [Project Page](https://qianmingduowan.github.io/Sat3DGen_project_page/)
- 💻 [GitHub](https://github.com/qianmingduowan/Sat3DGen)
- 🤗 [Model](https://huggingface.co/qian43/Sat3DGen)
- [Arxiv](arxiv.org/abs/2605.14984)

### Features

- **3D Mesh Generation**: Upload a satellite image → get a downloadable `.obj` mesh with in-browser 3D preview.
- **Video Rendering**: Select a satellite image and sky panorama → render a walkthrough video along a trajectory.

### Usage

Upload a satellite image or select one from the examples, then click "Generate".
The model weights are loaded automatically from HuggingFace.

### Input Requirements

The satellite image should be at **zoom level 20** (same as the [VIGOR](https://github.com/Jeff-Zilence/VIGOR) dataset), resized to **512×512** pixels. You can download satellite tiles at this zoom level from any map tile API (e.g. Google Maps, Bing Maps, Mapbox).

> **Note**: GPU hardware is recommended for reasonable inference speed.