index.html

<!DOCTYPE HTML>
<html lang="en"><head><meta http-equiv="Content-Type" content="text/html; charset=UTF-8">

<!-- Global site tag (gtag.js) - Google Analytics -->
<script async src="https://www.googletagmanager.com/gtag/js?id=G-3JLBSZZEVM"></script>
<script>
  window.dataLayer = window.dataLayer || [];
  function gtag(){dataLayer.push(arguments);}
  gtag('js', new Date());

  gtag('config', 'G-T9MWK0XS05');
</script>

<title>Simon Weber · Computer Vision &amp; Geometry</title>
<meta name="description" content="Simon Weber, Research Associate at the University of Oxford. Research in 3D computer vision, optimisation, and beyond-Euclidean geometry.">
  
<meta name="author" content="Simon Weber">
<meta name="viewport" content="width=device-width, initial-scale=1">

<link rel="stylesheet" type="text/css" href="stylesheet.css?v=3e040926">

</head>

<body>
<a class="skip-link" href="#main">Skip to content</a>
<div class="site-shell">
<header class="site-header"><nav aria-label="Main navigation"><a href="#about">About</a><a href="#news">News</a><a href="#preprints">Preprints</a><a href="#publications">Publications</a></nav></header>
<main id="main">
<section class="intro" id="about" aria-labelledby="name">
<div class="intro-copy"><p class="eyebrow">Computer vision · Geometry · Optimisation</p>
<h1 id="name">Simon Weber<span>.</span></h1>
<p class="role">Research Associate · University of Oxford</p>
<p class="research-statement">I study geometry and optimisation at the intersection of 3D computer vision and artificial intelligence.</p>
<div class="social-links" aria-label="Contact and profiles">
<a class="contact-link" href="mailto:simon.weber@cs.ox.ac.uk">Get in touch <span aria-hidden="true">↗</span></a>
<a href="https://scholar.google.com/citations?user=ENAQrtQAAAAJ&amp;hl">Google Scholar</a>
<a href="https://pixl.cs.ox.ac.uk">Group website</a>
<a href="https://x.com/SimWeberTUM">X</a>
<a href="https://www.linkedin.com/in/simon-weber-5b2896102/">LinkedIn</a>
</div></div>
<div class="portrait-frame"><img class="portrait" src="images/portrait_simon.png" alt="Simon Weber" width="458" height="684" fetchpriority="high"><span class="portrait-caption">Research Associate<br>University of Oxford</span></div>
</section>
<section class="background" aria-label="Research background"><p>I am a Research Associate at the University of Oxford in the <a href="https://pixl.cs.ox.ac.uk">PIXL lab</a> headed by <a href="https://ronnie-clark.co.uk">Ronald Clark</a> and an <a href="https://ellis.eu">ELLIS member</a>. I received my PhD in Computer Vision from the Technical University of Munich in the <a href="https://cvg.cit.tum.de"> Computer Vision Group</a>, headed by <a href="https://cvg.cit.tum.de/members/cremers"> Daniel Cremers </a>. I visited the <a href="https://www.imperial.ac.uk/dyson-robotics-lab/"> Dyson Robotics Lab </a> at Imperial College London, under the guidance of <a href="https://www.doc.ic.ac.uk/~ajd/"> Andrew J. Davison </a>. Before that, I studied at École normale supérieure (ENS Ulm) in Paris.</p><p>My research interests include 3D reconstruction for structure-from-motion and SLAM systems, beyond-Euclidean geometry, and semantic segmentation.</p></section>
<section class="news section" id="news" aria-labelledby="news-title"><div class="section-heading"><h2 id="news-title">Latest news</h2><span>A few recent updates</span></div>
<ul class="news-list"><li><time datetime="2026-09">09/2026</time><span>Our paper on <a href="https://arxiv.org/abs/2603.11396">asymmetric manifold learning</a> has been accepted at NeurIPS 2026!</span></li><li><time datetime="2026-04">04/2026</time><span>Our paper on <a href="https://arxiv.org/abs/2605.31559">rethinking attention</a> has been accepted at ICML 2026!</span></li><li><time datetime="2026-03">03/2026</time><span>I became an <a href="https://ellis.eu">ELLIS member</a>.</span></li></ul>
<details class="news-archive"><summary>Older news <span class="archive-count">10 updates</span></summary><ul class="news-list"><li><time datetime="2026-02">02/2026</time><span>I joined the University of Oxford, <a href="https://pixl.cs.ox.ac.uk">PIXL lab</a>!</span></li><li><time datetime="2025-11">11/2025</time><span>I successfully defended my PhD in computer vision.</span></li><li><time datetime="2025-06">06/2025</time><span>Visiting <a href="https://www.imperial.ac.uk/dyson-robotics-lab/">Andrew Davison's lab</a> at Imperial College London for a few months!</span></li><li><time datetime="2025-03">03/2025</time><span>Our <a href="https://arxiv.org/abs/2503.18010">Finsler Multi-Dimensional Scaling</a> paper has been accepted at CVPR 2025!</span></li><li><time datetime="2024-09">09/2024</time><span>Our CVPR paper <a href="https://arxiv.org/abs/2404.03778">Flattening the Parent Bias: Hierarchical Semantic Segmentation in the Poincaré Ball</a> is presented at the workshop <a href="https://sites.google.com/view/beyondeuclidean/home">Beyond Euclidean</a> at ECCV 2024.</span></li><li><time datetime="2024-07">07/2024</time><span>Our <a href="https://arxiv.org/abs/2405.05079">Power Variable Projection for Initialization-Free Large-Scale Bundle Adjustment</a> paper has been accepted at ECCV 2024.</span></li><li><time datetime="2024-07">07/2024</time><span>Visited <a href="https://iplab.dmi.unict.it/icvss2024/">International Computer Vision Summer School</a> in Sicily.</span></li><li><time datetime="2024-03">03/2024</time><span>Our <a href="https://arxiv.org/abs/2404.03778">Flattening the Parent Bias: Hierarchical Semantic Segmentation in the Poincaré Ball</a>  paper has been accepted at CVPR 2024.</span></li><li><time datetime="2024-03">03/2024</time><span>Our <a href="https://arxiv.org/abs/2404.03999">Finsler-Laplace-Beltrami Operators with Application to Shape Analysis</a>  paper has been accepted at CVPR 2024.</span></li><li><time datetime="2023-03">03/2023</time><span>Our <a href="https://arxiv.org/abs/2204.12834">Power Bundle Adjustment for Large-Scale 3D Reconstruction</a>  paper has been accepted at CVPR 2023.</span></li></ul></details></section>
<section class="section" id="preprints" aria-labelledby="preprints-title">
<div class="section-heading"><h2 id="preprints-title">Preprints</h2><span>2026</span></div>
<p class="section-note">* Equal contribution</p>
<article class="publication preprint">
<a class="paper-figure" href="https://arxiv.org/abs/2609.10531" aria-label="Read Guiding Image-to-3D Generation with Test-Time Partial Observations"><img src="images/chen2026guiding.png" alt="Figure from Guiding Image-to-3D Generation with Test-Time Partial Observations" loading="lazy" width="220" height="180"></a>
<div class="paper-content"><span class="venue">arXiv preprint · <time datetime="2026-09-09">September 2026</time></span>
<h3><a href="https://arxiv.org/abs/2609.10531">Guiding Image-to-3D Generation with Test-Time Partial Observations</a></h3>
<p class="authors">Jerred Chen, <strong>Simon Weber</strong>, <a href="https://ronnie-clark.co.uk">Ronald Clark</a></p>
<p class="paper-summary">Guiding pretrained image-to-3D models with partial geometric observations at test time, without retraining or finetuning.</p>
<a class="paper-link" href="https://arxiv.org/abs/2609.10531">Paper <span aria-hidden="true">↗</span></a>
</div>
</article>
<article class="publication preprint">
<a class="paper-figure" href="https://arxiv.org/abs/2608.18028" aria-label="Read Initialization-Free Bundle Adjustment Revisited: A Controlled Experimental Study"><img src="images/weber2026initialization.png" alt="Figure from Initialization-Free Bundle Adjustment Revisited: A Controlled Experimental Study" loading="lazy" width="220" height="180"></a>
<div class="paper-content"><span class="venue">arXiv preprint · <time datetime="2026-08-18">August 2026</time></span>
<h3><a href="https://arxiv.org/abs/2608.18028">Initialization-Free Bundle Adjustment Revisited: A Controlled Experimental Study</a></h3>
<p class="authors"><strong>Simon Weber</strong>*, Mateo de Mayo*, <a href="https://sail.hanyang.ac.kr">Je Hyeong Hong</a>, Carl Olsson, <a href="https://cvg.cit.tum.de/members/cremers">Daniel Cremers</a>, <a href="https://ronnie-clark.co.uk">Ronald Clark</a></p>
<p class="paper-summary">A controlled study of why low optimisation error does not always yield accurate metric 3D reconstructions in initialization-free bundle adjustment.</p>
<a class="paper-link" href="https://arxiv.org/abs/2608.18028">Paper <span aria-hidden="true">↗</span></a>
</div>
</article>
</section>
<section class="section" id="publications" aria-labelledby="publications-title"><div class="section-heading"><h2 id="publications-title">Publications</h2><span>2021–2026</span></div><p class="section-note">* Equal contribution · † Project lead</p>
<article class="publication">
<a class="paper-figure" href="https://arxiv.org/abs/2603.11396" aria-label="Read Harnessing Data Asymmetry in Manifold Learning"><img src="images/dages2026harnessing_2.png" alt="Figure from Harnessing Data Asymmetry in Manifold Learning" loading="lazy" width="220" height="180"><img class="alternate" src="images/dages2026harnessing_1.png" alt="" loading="lazy" width="220" height="180"></a>
<div class="paper-content"><span class="venue">NeurIPS, 2026</span>
<h3><a href="https://arxiv.org/abs/2603.11396">Harnessing Data Asymmetry in Manifold Learning</a></h3>
<p class="authors"><a href="https://tommoo.github.io">Thomas Dagès</a>, <strong>Simon Weber</strong>, <a href="https://ron.cs.technion.ac.il">Ron Kimmel</a>, <a href="https://cvg.cit.tum.de/members/cremers">Daniel Cremers</a></p>
<p class="paper-summary">Exploring asymmetric relationships in data through manifold learning.</p>
<a class="paper-link" href="https://arxiv.org/abs/2603.11396">Paper <span aria-hidden="true">↗</span></a></div>
</article>
<article class="publication">
<a class="paper-figure" href="https://arxiv.org/abs/2605.31559" aria-label="Read Functional Attention: From Pairwise Affinities to Functional Correspondences"><img src="images/xiao2026functional_2.png" alt="Figure from Functional Attention: From Pairwise Affinities to Functional Correspondences" loading="lazy" width="220" height="180"><img class="alternate" src="images/xiao2026functional_1.png" alt="" loading="lazy" width="220" height="180"></a>
<div class="paper-content"><span class="venue">ICML, 2026</span>
<h3><a href="https://arxiv.org/abs/2605.31559">Functional Attention: From Pairwise Affinities to Functional Correspondences</a></h3>
<p class="authors"><a href="https://xjffff.github.io">Jiefang Xiao</a>, <a href="https://maolingao.github.io">Maolin Gao</a><sup>†</sup>, <strong>Simon Weber</strong><sup>†</sup>, <a href="https://www.guandaoyang.com">Guandao Yang</a>, <a href="https://cvg.cit.tum.de/members/cremers">Daniel Cremers</a></p>
<p class="paper-summary">Rethinking attention through functional correspondences beyond pairwise affinities.</p>
<a class="paper-link" href="https://arxiv.org/abs/2605.31559">Paper <span aria-hidden="true">↗</span></a></div>
</article>
<article class="publication">
<a class="paper-figure" href="https://mediatum.ub.tum.de/?id=1783856" aria-label="Read From Krylov Spaces to Power Series: Advancing Large-Scale Bundle Adjustment with and without Initialization"><img src="images/Tum_logo_1.png" alt="Figure from From Krylov Spaces to Power Series: Advancing Large-Scale Bundle Adjustment with and without Initialization" loading="lazy" width="220" height="180"></a>
<div class="paper-content"><span class="venue">PhD Dissertation, Technical University of Munich, 2025</span>
<h3><a href="https://mediatum.ub.tum.de/?id=1783856">From Krylov Spaces to Power Series: Advancing Large-Scale Bundle Adjustment with and without Initialization</a></h3>
<p class="authors"><strong>Simon Weber</strong></p>
<p class="paper-summary">Methods for large-scale bundle adjustment, with and without an initial estimate.</p>
<a class="paper-link" href="https://mediatum.ub.tum.de/?id=1783856">Dissertation <span aria-hidden="true">↗</span></a></div>
</article>
<article class="publication">
<a class="paper-figure" href="https://arxiv.org/abs/2503.18010" aria-label="Read Finsler Multi-Dimensional Scaling: Manifold Learning for Asymmetric Dimensionality Reduction and Embedding"><img src="images/dages2025finsler_2.png" alt="Figure from Finsler Multi-Dimensional Scaling: Manifold Learning for Asymmetric Dimensionality Reduction and Embedding" loading="lazy" width="220" height="180"><img class="alternate" src="images/dages2025finsler_1.png" alt="" loading="lazy" width="220" height="180"></a>
<div class="paper-content"><span class="venue">CVPR, 2025</span>
<h3><a href="https://arxiv.org/abs/2503.18010">Finsler Multi-Dimensional Scaling: Manifold Learning for Asymmetric Dimensionality Reduction and Embedding</a></h3>
<p class="authors"><a href="https://tommoo.github.io">Thomas Dagès</a>*, <strong>Simon Weber</strong>*, <a href="https://www.ya-wei-eileen-lin.com">Ya-Wei Eileen Lin</a>, <a href="https://ronentalmon.com">Ronen Talmon</a>, <a href="https://cvg.cit.tum.de/members/cremers">Daniel Cremers</a>, <a href="https://mic.net.technion.ac.il">Michael Lindenbaum</a>, <a href="https://freddy.cs.technion.ac.il">Alfred M. Bruckstein</a>, <a href="https://ron.cs.technion.ac.il">Ron Kimmel</a></p>
<p class="paper-summary">Using Finsler geometry to embed data with asymmetric relationships.</p>
<a class="paper-link" href="https://arxiv.org/abs/2503.18010">Paper <span aria-hidden="true">↗</span></a></div>
</article>
<article class="publication">
<a class="paper-figure" href="https://arxiv.org/abs/2405.05079" aria-label="Read Power Variable Projection for Initialization-Free Large-Scale Bundle Adjustment"><img src="images/weber2024power_2.png" alt="Figure from Power Variable Projection for Initialization-Free Large-Scale Bundle Adjustment" loading="lazy" width="220" height="180"><img class="alternate" src="images/weber2024power_1.png" alt="" loading="lazy" width="220" height="180"></a>
<div class="paper-content"><span class="venue">ECCV, 2024</span>
<h3><a href="https://arxiv.org/abs/2405.05079">Power Variable Projection for Initialization-Free Large-Scale Bundle Adjustment</a></h3>
<p class="authors"><strong>Simon Weber</strong>, <a href="https://sail.hanyang.ac.kr">Je Hyeong Hong</a>, <a href="https://cvg.cit.tum.de/members/cremers">Daniel Cremers</a></p>
<p class="paper-summary">Large-scale bundle adjustment without requiring an initial reconstruction.</p>
<a class="paper-link" href="https://arxiv.org/abs/2405.05079">Paper <span aria-hidden="true">↗</span></a></div>
</article>
<article class="publication">
<a class="paper-figure" href="https://arxiv.org/abs/2404.03778" aria-label="Read Flattening the Parent Bias: Hierarchical Semantic Segmentation in the Poincaré Ball"><img src="images/weber2024flattening_2.png" alt="Figure from Flattening the Parent Bias: Hierarchical Semantic Segmentation in the Poincaré Ball" loading="lazy" width="220" height="180"><img class="alternate" src="images/weber2024flattening_1.png" alt="" loading="lazy" width="220" height="180"></a>
<div class="paper-content"><span class="venue">CVPR, 2024</span>
<h3><a href="https://arxiv.org/abs/2404.03778">Flattening the Parent Bias: Hierarchical Semantic Segmentation in the Poincaré Ball</a></h3>
<p class="authors"><strong>Simon Weber</strong>, Barış Zöngür, <a href="https://arnike.github.io">Nikita Araslanov</a>, <a href="https://cvg.cit.tum.de/members/cremers">Daniel Cremers</a></p>
<p class="paper-summary">Addressing parent bias in hierarchical semantic segmentation with hyperbolic geometry.</p>
<a class="paper-link" href="https://arxiv.org/abs/2404.03778">Paper <span aria-hidden="true">↗</span></a></div>
</article>
<article class="publication">
<a class="paper-figure" href="https://arxiv.org/abs/2404.03999" aria-label="Read Finsler-Laplace-Beltrami Operators with Application to Shape Analysis"><img src="images/weber2024finsler_2.png" alt="Figure from Finsler-Laplace-Beltrami Operators with Application to Shape Analysis" loading="lazy" width="220" height="180"><img class="alternate" src="images/weber2024finsler_1.png" alt="" loading="lazy" width="220" height="180"></a>
<div class="paper-content"><span class="venue">CVPR, 2024</span>
<h3><a href="https://arxiv.org/abs/2404.03999">Finsler-Laplace-Beltrami Operators with Application to Shape Analysis</a></h3>
<p class="authors"><strong>Simon Weber</strong>*, <a href="https://tommoo.github.io">Thomas Dagès</a>*, <a href="https://maolingao.github.io">Maolin Gao</a>, <a href="https://cvg.cit.tum.de/members/cremers">Daniel Cremers</a></p>
<p class="paper-summary">Bringing Finsler geometry to differential operators for shape analysis.</p>
<a class="paper-link" href="https://arxiv.org/abs/2404.03999">Paper <span aria-hidden="true">↗</span></a></div>
</article>
<article class="publication">
<a class="paper-figure" href="https://arxiv.org/abs/2204.12834" aria-label="Read Power Bundle Adjustment for Large-Scale 3D Reconstruction"><img src="images/weber2023power_2.png" alt="Figure from Power Bundle Adjustment for Large-Scale 3D Reconstruction" loading="lazy" width="220" height="180"><img class="alternate" src="images/weber2023power_1.png" alt="" loading="lazy" width="220" height="180"></a>
<div class="paper-content"><span class="venue">CVPR, 2023</span>
<h3><a href="https://arxiv.org/abs/2204.12834">Power Bundle Adjustment for Large-Scale 3D Reconstruction</a></h3>
<p class="authors"><strong>Simon Weber</strong>, <a href="https://cvg.cit.tum.de/members/demmeln">Nikolaus Demmel</a>, Tin Chon Chan, <a href="https://cvg.cit.tum.de/members/cremers">Daniel Cremers</a></p>
<p class="paper-summary">A power-series approach to bundle adjustment for large-scale 3D reconstruction.</p>
<a class="paper-link" href="https://arxiv.org/abs/2204.12834">Paper <span aria-hidden="true">↗</span></a></div>
</article>
<article class="publication">
<a class="paper-figure" href="https://arxiv.org/abs/2110.04015" aria-label="Read Multidirectional Conjugate Gradients for Scalable Bundle Adjustment"><img src="images/weber2021multidirectional_2.png" alt="Figure from Multidirectional Conjugate Gradients for Scalable Bundle Adjustment" loading="lazy" width="220" height="180"><img class="alternate" src="images/weber2021multidirectional_1.png" alt="" loading="lazy" width="220" height="180"></a>
<div class="paper-content"><span class="venue">GCPR, 2021</span>
<h3><a href="https://arxiv.org/abs/2110.04015">Multidirectional Conjugate Gradients for Scalable Bundle Adjustment</a></h3>
<p class="authors"><strong>Simon Weber</strong>, <a href="https://cvg.cit.tum.de/members/demmeln">Nikolaus Demmel</a>, <a href="https://cvg.cit.tum.de/members/cremers">Daniel Cremers</a></p>
<p class="paper-summary">Exploring multidirectional conjugate gradients for scalable bundle adjustment.</p>
<a class="paper-link" href="https://arxiv.org/abs/2110.04015">Paper <span aria-hidden="true">↗</span></a></div>
</article>
</section></main>
<footer><span>Simon Weber · University of Oxford</span><span>Inspired by <a href="https://jonbarron.info/">Jon Barron’s template</a> · <a href="https://github.com/jonbarron/jonbarron_website">Original source</a></span></footer>
</div></body></html>