Unsupervised 3D Occupancy Learning — Poster
A conference poster detailing a method for unsupervised 3D occupancy learning from sparse point clouds, featuring mathematical formulations, qualitative comparisons on SRB and human datasets, and quantitative results on FAUST and ShapeNet.
论文上下文
Paper title: Unsupervised Occupancy Learning from Sparse Point Cloud (CVPR 2024) Abstract: A conference poster detailing a method for unsupervised 3D occupancy learning from sparse point clouds, featuring mathematical formulations, qualitative comparisons on SRB and human datasets, and quantitative results on FAUST and ShapeNet. Paper body (method & results): <!DOCTYPE html> <html lang="en"> <head> <meta content="text/html; charset=utf-8" http-equiv="content-type"/> <title>Unsupervised Occupancy Learning from Sparse Point Cloud</title> <!--Generated on Thu May 2 16:43:30 2024 by LaTeXML (version 0.8.8) http://dlmf.nist.gov/LaTeXML/.--> <meta content="width=device-width, initial-scale=1, shrink-to-fit=no" name="viewport"/> <link href="https://cdn.jsdelivr.net/npm/bootstrap@5.3.0/dist/css/bootstrap.min.css" rel="stylesheet" type="text/css"/> <link href="/static/browse/0.3.4/css/ar5iv.0.7.9.min.css" rel="stylesheet" type="text/css"/> <link href="/static/browse/0.3.4/css/ar5iv-fonts.0.7.9.min.css" rel="stylesheet" type="text/css"/> <link href="/static/browse/0.3.4/css/latexml_styles.css" rel="stylesheet" type="text/css"/> <script src="https://cdn.jsdelivr.net/npm/bootstrap@5.3.0/dist/js/bootstrap.bundle.min.js"></script> <script src="https://cdnjs.cloudflare.com/ajax/libs/html2canvas/1.3.3/html2canvas.min.js"></script> <script src="/static/browse/0.3.4/js/addons_new.js"></script> <script src="/static/browse/0.3.4/js/feedbackOverlay.js"></script> <base href="/html/2404.02759v1/"/></head> <body> <nav class="ltx_page_navbar"> <nav class="ltx_TOC"> <ol class="ltx_toclist"> <li class="ltx_tocentry ltx_tocentry_section"><a class="ltx_ref" href="https://arxiv.org/html/2404.02759v1#S1" title="In Unsupervised Occupancy Learning from Sparse Point Cloud"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">1 </span>Introduction</span></a></li> <li class="ltx_tocentry ltx_tocentry_section"><a class="ltx_ref" href="https://arxiv.org/html/2404.02759v1#S2" title="In Unsupervised Occupancy Learning from Sparse Point Cloud"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">2 </span>Related Work</span></a></li> <li class="ltx_tocentry ltx_tocentry_section"> <a class="ltx_ref" href="https://arxiv.org/html/2404.02759v1#S3" title="In Unsupervised Occupancy Learning from Sparse Point Cloud"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3 </span>Method</span></a> <ol class="ltx_toclist ltx_toclist_section"> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2404.02759v1#S3.SS1" title="In 3 Method ‣ Unsupervised Occupancy Learning from Sparse Point Cloud"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3.1 </span>Learning Occupancy Through Margin Uncertainty Sampling</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2404.02759v1#S3.SS2" title="In 3 Method ‣ Unsupervised Occupancy Learning from Sparse Point Cloud"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3.2 </span>Entropy Based Regularization</span></a></li> </ol> </li> <li class="ltx_tocentry ltx_tocentry_section"><a class="ltx_ref" href="https://arxiv.org/html/2404.02759v1#S4" title="In Unsupervised Occupancy Learning from Sparse Point Cloud"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">4 </span>Implementation Details</span></a></li> <li class="ltx_tocentry ltx_tocentry_section"> <a class="ltx_ref" href="https://arxiv.org/html/2404.02759v1#S5" title="In Unsupervised Occupancy Learning from Sparse Point Cloud"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">5 </span>Results</span></a> <ol class="ltx_toclist ltx_toclist_section"> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2404.02759v1#S5.SS1" title="In 5 Results ‣ Unsupervised Occupancy Learning from Sparse Point Cloud"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">5.1 </span>Metrics</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2404.02759v1#S5.SS2" title="In 5 Results ‣ Unsupervised Occupancy Learning from Sparse Point Cloud"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">5.2 </span>Datasets and Input Definitions</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2404.02759v1#S5.SS3" title="In 5 Results ‣ Unsupervised Occupancy Learning from Sparse Point Cloud"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">5.3 </span>Object Level Reconstruction</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2404.02759v1#S5.SS4" title="In 5 Results ‣ Unsupervised Occupancy Learning from Sparse Point Cloud"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">5.4 </span>Real Articulated Shape Reconstruction</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2404.02759v1#S5.SS5" title="In 5 Results ‣ Unsupervised Occupancy Learning from Sparse Point Cloud"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">5.5 </span>Real Scene Level Reconstruction</span></a></li> </ol> </li> <li class="ltx_tocentry ltx_tocentry_section"><a class="ltx_ref" href="https://arxiv.org/html/2404.02759v1#S6" title="In Unsupervised Occupancy Learning from Sparse Point Cloud"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">6 </span>Ablation Studies</span></a></li> <li class="ltx_tocentry ltx_tocentry_section"><a class="ltx_ref" href="https://arxiv.org/html/2404.02759v1#S7" title="In Unsupervised Occupancy Learning from Sparse Point Cloud"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">7 </span>Limitations</span></a></li> <li class="ltx_tocentry ltx_tocentry_section"><a class="ltx_ref" href="https://arxiv.org/html/2404.02759v1#S8" title="In Unsupervised Occupancy Learning from Sparse Point Cloud"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">8 </span>Conclusion</span></a></li> <li class="ltx_tocentry ltx_tocentry_appendix"><a class="ltx_ref" href="https://arxiv.org/html/2404.02759v1#A1" title="In Unsupervised Occupancy Learning from Sparse Point Cloud"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref"><span class="ltx_text" style="font-size:144%;">A</span> </span><span class="ltx_text" style="font-size:144%;">Additional Implementation Details</span></span></a></li> <li class="ltx_tocentry ltx_tocentry_appendix"> <a class="ltx_ref" href="https://arxiv.org/html/2404.02759v1#A2" title="In Unsupervised Occupancy Learning from Sparse Point Cloud"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref"><span class="ltx_text" style="font-size:144%;">B</span> </span><span class="ltx_text" style="font-size:144%;">Metrics</span></span></a> <ol class="ltx_toclist ltx_toclist_appendix"> <li class="ltx_tocentry ltx_tocentry_paragraph"><a class="ltx_ref" href="https://arxiv.org/html/2404.02759v1#A2.SS0.SSS0.Px1" title="In Appendix B Metrics ‣ Unsupervised Occupancy Learning from Sparse Point Cloud"><span class="ltx_text ltx_ref_title">Chamfer Distance (CD<sub class="ltx_sub">1</sub>)</span></a></li> <li class="ltx_tocentry ltx_tocentry_paragraph"><a class="ltx_ref" href="https://arxiv.org/html/2404.02759v1#A2.SS0.SSS0.Px2" title="In Appendix B Metrics ‣ Unsupervised Occupancy Learning from Sparse Point Cloud"><span class="ltx_text ltx_ref_title">Chamfer Distance (CD<sub class="ltx_sub">2</sub>)</span></a></li> <li class="ltx_tocentry ltx_tocentry_paragraph"><a class="ltx_ref" href="https://arxiv.org/html/2404.02759v1#A2.SS0.SSS0.Px3" title="In Appendix B Metrics ‣ Unsupervised Occupancy Learning from Sparse Point Cloud"><span class="ltx_text ltx_ref_title">F-Score (FS)</span></a></li> <li class="ltx_tocentry ltx_tocentry_paragraph"><a class="ltx_ref" href="https://arxiv.org/html/2404.02759v1#A2.SS0.SSS0.Px4" title="In Appendix B Metrics ‣ Unsupervised Occupancy Learning from Sparse Point Cloud"><span class="ltx_text ltx_ref_title">Normal consistency (NC)</span></a></li> <li class="ltx_tocentry ltx_tocentry_paragraph"><a class="ltx_ref" href="https://arxiv.org/html/2404.02759v1#A2.SS0.SSS0.Px5" title="In Appendix B Metrics ‣ Unsupervised Occupancy Learning from Sparse Point Cloud"><span class="ltx_text ltx_ref_title">Hausdorff distance (HD)</span></a></li> </ol> </li> </ol></nav> </nav> <div class="ltx_page_main"> <div class="ltx_page_content"> <article class="ltx_document ltx_authors_1line"> <h1 class="ltx_title ltx_title_document">Unsupervised Occupancy Learning from Sparse Point Cloud</h1> <div class="ltx_authors"> <span class="ltx_creator ltx_role_author"> <span class="ltx_personname">Amine Ouasfi Adnane Boukhayma <br class="ltx_break"/>Inria, Univ. Rennes, CNRS, IRISA, M2S, France </span></span> </div> <div class="ltx_abstract"> <h6 class="ltx_title ltx_title_abstract">Abstract</h6> <p class="ltx_p" id="id1.id1">Implicit Neural Representations have gained prominence as a powerful framework for capturing complex data modalities, encompassing a wide range from 3D shapes to images and audio. Within the realm of 3D shape representation, Neural Signed Distance Functions (SDF) have demonstrated remarkable potential in faithfully encoding intricate shape geometry. However, learning SDFs from 3D point clouds in the absence of ground truth supervision remains a very challenging task. In this paper, we propose a method to infer occupancy fields instead of SDFs as they are easier to learn from sparse inputs. We leverage a margin-based uncertainty measure to differentiably sample from the decision boundary of the occupancy function and supervise the sampled boundary points using the input point cloud. We further stabilise the optimization process at the early stages of the training by biasing the occupancy function towards minimal entropy fields while maximizing its entropy at the input point cloud. Through extensive experiments and evaluations, we illustrate the efficacy of our proposed method, highlighting its capacity to improve implicit shape inference with respect to baselines and the state-of-the-art using synthetic and real data.</p> </div> <section class="ltx_section" id="S1"> <h2 class="ltx_title ltx_title_section"> <span class="ltx_tag ltx_tag_section">1 </span>Introduction</h2> <div class="ltx_para" id="S1.p1"> <p class="ltx_p" id="S1.p1.1">Capturing full 3D shape from limited and corrupted data is a long standing problem. In this regard, one of the data type instances that has received increasing attention and investigation from computer vision, graphics and machine learning is point clouds. This interest emanates primarily from the ubiquity of this light, although topologically incomplete 3D representation, either as acquired <em class="ltx_emph ltx_font_italic" id="S1.p1.1.1">e.g</em>.<span class="ltx_text" id="S1.p1.1.2"></span> through industrial and commodity depth sensors, or as an intermediate representation within computational photogrammetry <cite class="ltx_cite ltx_citemacro_cite">[<a class="ltx_ref" href="https://arxiv.or
