---
license: apache-2.0
pipeline_tag: image-to-3d
---
<p align="center">
  <h1 align="center">PointDiT: Pixel-Space Diffusion for Monocular Geometry Estimation</h1>
  <p align="center">
    <a href="https://haofeixu.github.io/">Haofei Xu</a><sup>1,2,3</sup> &middot;
    <a href="https://rundiwu.github.io/">Rundi Wu</a><sup>1</sup> &middot;
    <a href="https://henzler.github.io/">Philipp Henzler</a><sup>1</sup> &middot;
    <a href="https://scholar.google.com/citations?user=XwzlnZoAAAAJ">Nikolai Kalischek</a><sup>1</sup> &middot;
    <a href="https://scholar.google.com/citations?user=vMcsUNAAAAAJ">Michael Oechsle</a><sup>1</sup>
    <br>
    <a href="https://scholar.google.com/citations?user=bERItx8AAAAJ">Fabian Manhardt</a><sup>1</sup> &middot;
    <a href="https://people.inf.ethz.ch/marc.pollefeys/">Marc Pollefeys</a><sup>2,4</sup> &middot;
    <a href="http://www.cvlibs.net/">Andreas Geiger</a><sup>3,5</sup> &middot;
    <a href="https://federicotombari.github.io/">Federico Tombari</a><sup>1,6</sup> &middot;
    <a href="https://m-niemeyer.github.io/">Michael Niemeyer</a><sup>1</sup>
  </p>
  <p align="center">
    <sup>1</sup>Google &nbsp; <sup>2</sup>ETH Zurich &nbsp; <sup>3</sup>University of Tübingen, Tübingen AI Center &nbsp; <sup>4</sup>Microsoft &nbsp; <sup>5</sup>KE:SAI &nbsp; <sup>6</sup>TUM
  </p>
  <h3 align="center">ICML 2026</h3>
  <h3 align="center">
    <a href="https://arxiv.org/abs/2607.02515">Paper</a> | <a href="https://haofeixu.github.io/pointdit/">Project Page</a>  | <a href="https://github.com/google-research/pointdit">Code</a> | <a href="https://github.com/google-research/pointdit/blob/main/MODELS.md">Models Details</a>
  </h3>
</p>

<p align="center">
  <a href="https://haofeixu.github.io/assets/pointdit.mp4">
    <img src="https://github.com/google-research/pointdit/blob/main/assets/teaser.gif?raw=true" alt="PointDiT teaser" width="100%">
  </a>
</p>