1<!DOCTYPE html> 2<html> 3<head> 4 <meta charset="utf-8"> 5 <meta name="description" 6 content="Deformable Neural Radiance Fields creates free-viewpoint portraits (nerfies) from casually captured videos."> 7 <meta name="keywords" content="Nerfies, D-NeRF, NeRF"> 8 <meta name="viewport" content="width=device-width, initial-scale=1"> 9 <title>DINO-SLAM: DINO-informed RGB-D SLAM for Neural Implicit and Explicit Representations</title> 10 11 <!-- Global site tag (gtag.js) - Google Analytics -->
vendor: 66 bytes, lines 11-12
11 12 <script async src="https://www.googletagmanager.com/gtag/js?id=
12G-PYVRSFMDRL
vendor: 14 bytes, lines 12-13
12"></script> 13
13<script> 14
vendor: 155 bytes, lines 14-22
14window.dataLayer = window.dataLayer || []; 15 16 function gtag() { 17 dataLayer.push(arguments); 18 } 19 20 gtag('js', new Date()); 21 22 gtag('config', '
22G-PYVRSFMDRL
vendor: 6 bytes, lines 22-23
22'); 23
23</script>
23 24 25 <link href="https://fonts.googleapis.com/css?family=Google+Sans|Noto+Sans|Castoro" 26 rel="stylesheet"> 27 28 <link rel="stylesheet" href="./static/css/bulma.min.css"> 29 <link rel="stylesheet" href="./static/css/bulma-carousel.min.css"> 30 <link rel="stylesheet" href="./static/css/bulma-slider.min.css"> 31 <link rel="stylesheet" href="./static/css/fontawesome.all.min.css"> 32 <link rel="stylesheet" 33 href="https://cdn.jsdelivr.net/gh/jpswalsh/academicons@1/css/academicons.min.css"> 34 <link rel="stylesheet" href="./static/css/index.css"> 35 <link rel="icon" href="./static/images/robot.svg"> 36 37
37<script src="https://ajax.googleapis.com/ajax/libs/jquery/3.5.1/jquery.min.js"></script>
37 38
38<script defer src="./static/js/fontawesome.all.min.js"></script>
38 39
39<script src="./static/js/bulma-carousel.min.js"></script>
39 40
40<script src="./static/js/bulma-slider.min.js"></script>
40 41
41<script src="./static/js/index.js"></script>
41 42</head> 43<body> 44 45<nav class="navbar" role="navigation" aria-label="main navigation"> 46 <div class="navbar-brand"> 47 <a role="button" class="navbar-burger" aria-label="menu" aria-expanded="false"> 48 <span aria-hidden="true"></span> 49 <span aria-hidden="true"></span> 50 <span aria-hidden="true"></span> 51 </a> 52 </div> 53 54 55 </div> 56</nav> 57 58 59<section class="hero"> 60 <div class="hero-body"> 61 <div class="container is-max-desktop"> 62 <div class="columns is-centered"> 63 <div class="column has-text-centered"> 64 <h1 class="title is-1 publication-title">DINO-SLAM: DINO-informed RGB-D SLAM for Neural Implicit and Explicit Representations</h1> 65 <h1 class="title is-size-3" style="color:#5a6268;">ECCV 2026</h1> 66 <div class="is-size-5 publication-authors"> 67 <span class="author-block"> 68 <a href="https://zorangong.github.io/">Ziren Gong</a><sup>1</sup>,</span> 69 <span class="author-block"> 70 <a href="https://scholar.google.com/citations?hl=zh-CN&user=1P5NDzQAAAAJ">Xiaohan Li</a><sup>2</sup>, 71 </span> 72 <span class="author-block"> 73 <a href="https://fabiotosi92.github.io/">Fabio Tosi</a><sup>1</sup>, 74 </span> 75 <span class="author-block"> 76 <a href="https://scholar.google.com/citations?hl=zh-CN&user=H_uqMz0AAAAJ">Youmin Zhang</a><sup>3</sup>, 77 </span> 78 <span class="author-block"> 79 <a href="https://stefanomattoccia.github.io/">Stefano Mattocia</a><sup>1</sup>, 80 </span> 81 <span class="author-block"> 82 <a href="https://jianfei-cai.github.io/">Jun Wu</a><sup>4</sup>, 83 </span> 84 <span class="author-block"> 85 <a href="https://mattpoggi.github.io/">Matteo Poggi</a><sup>1</sup>, 86 </span> 87 </div> 88 89 <div class="is-size-5 publication-authors"> 90 <span class="author-block"><sup>1</sup>University of Bologna,</span> 91 <span class="author-block"><sup>2</sup>The University of Hongkong</span> 92 <span class="author-block"><sup>3</sup>Rawmantic AI</span> 93 <span class="author-block"><sup>4</sup>Fudan University</span> 94 </div> 95 96 <div class="column has-text-centered"> 97 <div class="publication-links"> 98 <!-- PDF Link. --> 99 <span class="link-block"> 100 <a href="https://arxiv.org/abs/2507.19474" 101 class="external-link button is-normal is-rounded is-dark"> 102 <span class="icon"> 103 <i class="fas fa-file-pdf"></i> 104 </span> 105 <span>Paper</span> 106 </a> 107 </span> 108 <!-- Video Link. --> 109 <span class="link-block"> 110 <a href="" 111 class="external-link button is-normal is-rounded is-dark"> 112 <span class="icon"> 113 <i class="fab fa-youtube"></i> 114 </span> 115 <span>Video</span> 116 </a> 117 </span> 118 <!-- code Link. --> 119 <span class="link-block"> 120 <a href="" 121 class="external-link button is-normal is-rounded is-dark"> 122 <span class="icon"> 123 <i class="fab fa-github"></i> 124 </span> 125 <span>Code (coming soon)</span> 126 </a> 127 </span> 128 <!-- Poster Link. --> 129 <span class="link-block"> 130 <a href="" 131 class="external-link button is-normal is-rounded is-dark"> 132 <span class="icon"> 133 <i class="far fa-images"></i> 134 </span> 135 <span>Poster (coming soon)</span> 136 </a> 137 <!-- results Link. --> 138 <span class="link-block"> 139 <a href="" 140 class="external-link button is-normal is-rounded is-dark">
141 <span class="icon"> 142 <i class="far fa-images"></i> 143 </span> 144 <span>Results</span> 145 </a> 146 </div> 147 148 </div> 149 </div> 150 </div> 151 </div> 152 </div> 153</section> 154 155<section class="hero teaser"> 156 <div class="container is-max-desktop"> 157 <div class="hero-body"> 158 <img style="width: 100%;" src="./static/images/overview.png" 159 alt="NIS-SLAM architecture."/> 160 <h2 class="subtitle has-text-justified"> 161 <p> 162 <b>Pipeline of DINO </b>. DINO-SLAM is a geometry-aware visual foundation feature informed neural SLAM framework. Both neural representations share the same structure of Scene Geometry Encoder (SGE) to capture geometry-aware DINO (geoDINO) and DINO features. 163 </p> 164 </h2> 165 </div> 166 </div> 167</section> 168 169<section class="section"> 170 <div class="container is-max-desktop"> 171 <!-- Abstract. --> 172 <div class="columns is-centered has-text-centered"> 173 <div class="column is-four-fifths"> 174 <h2 class="title is-3">Abstract</h2> 175 <h2 class="content has-text-justified"> 176 <p> 177 a DINO-informed design strategy to enhance implicit (Neural Radiance Field -- NeRF) and explicit representations (Gaussian Splatting -- GS) in SLAM systems through the more comprehensive semantics understanding enabled by DINO. 178 </p> 179 <p> 180 This latter alone, however, lacks proper 3D geometry understanding, allowing only for marginal improvements. Therefore, we rely on a Scene Geometry Encoder (SGE) to enrich DINO features into geometry-aware DINO features (geoDINO), to better understand those geometric relationships that vanilla DINO features fail to capture. Building upon it, we propose two foundational paradigms for NeRF and GS SLAM systems integrating geoDINO features. 181 </p> 182 <p> 183 Compared to state-of-the-art methods, our DINO-informed pipelines achieve superior performance on the Replica, ScanNet, and TUM datasets. 184 </p> 185 </h2> 186 </div> 187 </div> 188 <!--/ Abstract. --> 189 190 <!-- Paper Framework --> 191 <div class="columns is-centered has-text-centered"> 192 <div class="column is-four-fifths"> 193 <h2 class="title is-3">Framework</h2> 194 <div class="content has-text-centered"> 195 <img style="width: 100%;" src="./static/images/DINO_extractor.png" 196 alt="Reconstruction Result."/> 197 </div> 198 <p> 199 <strong><font size="4">Scene Geometry Encoder (SGE):</font></strong> Our SGE has two outputs: 1) DINO features, simply extracted from RGB frames by the DINO model, 200 2) geoDINO features obtained from the joint processing of DINO features and depth maps, projected respectively into appearance and geometric features, and used as keys and queries to attend over. 201 </p> 202 203 <div class="content has-text-centered"> 204 <img style="width: 100%;" src="./static/images/gs_nerf_pipeline.png" 205 alt="Reconstruction Result."/> 206 </div> 207 <p> 208 <strong><font size="4">Framework of Nueral Implicit and Explicit Representations:</font></strong> In the NeRF-based pipeline (left), we leverage EDINO features to supervise the tri-plane (red double arrow) while DINO features guide the optimization of the estimated DINO feature map (red double arrow). 209 In the 3DGS-based pipeline (right), we incorporate the corresponding EDINO feature into the parameters of each Gaussian (blue arrow). Meanwhile, the DINO features are leveraged to guide the learning of DINO information within the representation (blue double arrow). 210 </p> 211 </h2> 212 </div> 213 </div> 214 <!--/ Paper Framework. --> 215 </div> 216</section> 217 218 219<section class="section"> 220 <div class="container is-max-desktop"> 221 <!-- Visualization --> 222 <div class="columns is-centered"> 223 <div class="column is-full-width"> 224 <h2 class="title is-3">Visualization</h2> 225 <div class="has-text-centered"> 226 <img style="width: 100%;" src="./static/images/vis4compar.png" 227 alt="Reconstruction Result."/> 228 </div> 229 <h2 class="content has-text-justified"> 230 <p> 231 <strong><font size="4">Visualization of DINO-SLAM and baselines on the Replica dataset.:</font></strong> We present details of reconstruction and rendering quality with red boxes and blue boxes. The two left columns present the reconstruction performance of our DINO-SLAM (NeRF). Our method yields superior mesh results with detailed textures and better completeness. 232 The two right columns show the rendering results of our DINO-SLAM (GS). Our DINO-SLAM demonstrates high-fidelity and realistic images, especially on text tags and object textures. 233 </p> 234 </h2> 235 <div class="has-text-centered"> 236 <img style="width: 100%;" src="./static/images/teaser.png" 237 alt="Reconstruction Result."/> 238 </div> 239 <h2 class="content has-text-justified"> 240 <p> 241 <strong><font size="4">More results and performance on Replica.</font></strong> DINO-SLAM achieves superior camera tracking (left, evaluated on scannet) and 3D reconstruction quality (middle, evaluated on Replica). 242 Our approach is general: it can be integrated into existing NeRF/GS SLAM systems and boost their results (right, showing some representative GS methods being improved by our proposal).
243 </p> 244 </h2> 245 246 247 </div> 248 </div> 249 <!--/ Visualization --> 250 251 <!-- Results --> 252 <div class="columns has-text-justified"> 253 <div class="column is-full-width"> 254 <h2 class="title is-3">Reconstruction & Segmentation Results</h2> 255 <p> 256 Our final reconstruction and segmentation results are available now. You can find it in the above <i>Results</i> button or you can download it through <a href="">Google Drive</a>. 257 </p> 258 </div> 259 </div> 260 261 <!-- Concurrent Work. --> 262 <div class="columns is-centered"> 263 <div class="column is-full-width"> 264 <h2 class="title is-3">Related Links</h2> 265 266 <div class="content has-text-justified"> 267 <p> 268 There's a lot of excellent work that was introduced around the same time as ours. 269 </p> 270 <p> 271 <a href="https://arxiv.org/abs/2507.19474">DINO-SLAM: DINO-informed RGB-D SLAM for Neural Implicit and Explicit Representations</a> provides two foundational paradigms for NeRF and 3DGS SLAM systems integrating geometry-aware DINO features. 272 </p> 273 <p> 274 <a href="https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=11127551">HS-SLAM: Hybrid Representation with Structural Supervision for Improved Dense SLAM</a> introduce HS-SLAM to enhance scene representations, capture structural information, and maintain global consistency. 275 </p> 276 <p> 277 <a href="https://arxiv.org/abs/2402.13255">How NeRFs and 3D Gaussian Splatting are Reshaping SLAM: a Survey</a> provides the first comprehensive survey of SLAM progress through the lens of the latest advancements in radiance fields (NeRF and 3DGS). 278 </p> 279 <p> 280 <a href="https://github.com/youmi-zym/GO-SLAM">Go-SLAM</a> introduces a deep-learning-based dense visual SLAM framework that achieves real-time global optimization of poses and 3D reconstruction. 281 </p> 282 </div> 283 </div> 284 </div> 285 <!--/ Concurrent Work. --> 286 287 </div> 288</section> 289 290 291<section class="section" id="BibTeX"> 292 <div class="container is-max-desktop content"> 293 <h2 class="title">BibTeX</h2> 294 <pre><code>@article{gong2025dino, 295 title={DINO-SLAM: DINO-informed RGB-D SLAM for Neural Implicit and Explicit Representations}, 296 author={Gong, Ziren and Li, Xiaohan and Tosi, Fabio and Zhang, Youmin and Mattoccia, Stefano and Wu, Jun and Poggi, Matteo}, 297 journal={arXiv preprint arXiv:2507.19474}, 298 year={2025} 299}</code></pre> 300 </div> 301</section> 302 303 304<footer class="footer"> 305 <div class="container"> 306 <div class="content has-text-centered"> 307 <a class="icon-link" 308 href="./static/videos/nerfies_paper.pdf"> 309 <i class="fas fa-file-pdf"></i> 310 </a> 311 <a class="icon-link" href="https://github.com/keunhong" class="external-link" disabled> 312 <i class="fab fa-github"></i> 313 </a> 314 </div> 315 <div class="columns is-centered"> 316 <div class="column is-8"> 317 <div class="content"> 318 <p> 319 This website is licensed under a <a rel="license" 320 href="http://creativecommons.org/licenses/by-sa/4.0/">Creative 321 Commons Attribution-ShareAlike 4.0 International License</a>. 322 </p> 323 <p> 324 This means you are free to borrow the <a 325 href="https://github.com/nerfies/nerfies.github.io">source code</a> of this website, 326 we just ask that you link back to this page in the footer. 327 Please remember to remove the analytics code included in the header of the website which 328 you do not want on your website. 329 </p> 330 </div> 331 </div> 332 </div> 333 </div> 334</footer> 335 336</body> 337</html>
Line numbers count LF bytes from the start of the resource, as the search results do. Vendor segments are library code the classifier recognised; they are stored but not indexed. Bytes are shown as Latin1 characters, one per byte.