{"id":14778,"date":"2026-01-04T15:14:01","date_gmt":"2026-01-04T14:14:01","guid":{"rendered":"https:\/\/www.keris-studio.fr\/blog\/?p=14778"},"modified":"2026-02-03T08:46:11","modified_gmt":"2026-02-03T07:46:11","slug":"gaussian-splatting-a-high-quality-3d-scene-representation","status":"publish","type":"post","link":"https:\/\/www.keris-studio.fr\/blog\/?p=14778","title":{"rendered":"Gaussian Splatting: A High-Quality 3D Scene Representation"},"content":{"rendered":"<h1>Creating a Gaussian Splatting workflow<\/h1>\n<p><strong>Gaussian Splatting<\/strong>\u00a0is a novel 3D scene representation technique that&rsquo;s revolutionizing novel view synthesis (creating new camera viewpoints of 3D scenes). It emerged as a state-of-the-art alternative to Neural Radiance Fields (NeRFs) in 2023.<\/p>\n<p><strong>Core Concept<\/strong><\/p>\n<p>Instead of representing scenes with neural networks (like NeRFs) or explicit meshes, Gaussian Splatting represents scenes as\u00a0<strong>millions of tiny 3D Gaussian \u00ab\u00a0blobs\u00a0\u00bb<\/strong>\u00a0that can be efficiently rendered using a technique called \u00ab\u00a0splatting.\u00a0\u00bb<\/p>\n<p>The technique represents a significant shift from implicit neural representations back toward explicit, differentiable graphics primitives, but with modern optimization techniques that achieve unprecedented quality and speed.<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" width=\"1443\" height=\"1030\" class=\"wp-image-14779\" src=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/d-laurent-downloads-splat-image-5-png.png\" alt=\"D:\\LAURENT\\Downloads\\splat-image (5).png\" srcset=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/d-laurent-downloads-splat-image-5-png.png 1443w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/d-laurent-downloads-splat-image-5-png-300x214.png 300w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/d-laurent-downloads-splat-image-5-png-1024x731.png 1024w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/d-laurent-downloads-splat-image-5-png-768x548.png 768w\" sizes=\"auto, (max-width: 1443px) 100vw, 1443px\" \/><!--more--><\/p>\n<h1>Step one<\/h1>\n<p>Choosing the platform\u00a0: NerfStudio https:\/\/docs.nerf.studio\/<\/p>\n<p><strong>Nerfstudio: A Framework for Neural Radiance Field (NeRF) Development<\/strong><\/p>\n<p><strong>Nerfstudio<\/strong>\u00a0is an open-source, modular framework designed to make\u00a0<strong>Neural Radiance Field (NeRF)<\/strong>\u00a0research and development more accessible, organized, and collaborative. Think of it as \u00ab\u00a0PyTorch Lightning for NeRFs\u00a0\u00bb or \u00ab\u00a0Hugging Face for 3D reconstruction.\u00a0\u00bb<\/p>\n<p><strong>Training is working as this:<\/strong><\/p>\n<ul>\n<li><strong>ns-train nerfacto &#8211;data data\/processed<\/strong><\/li>\n<li><strong>ns-train instant-ngp &#8211;data data\/processed # faster<\/strong><\/li>\n<li><strong>ns-train gaussian-splatting &#8211;data data\/processed # real-time rendering<\/strong><\/li>\n<\/ul>\n<h2>Installation: Mandatory Prerequisites<\/h2>\n<p>Before running the script, you must install these three tools. Without them, the script will fail 100% during compilation.<\/p>\n<h3>Visual Studio Build Tools (The C++ Compiler)<\/h3>\n<p>Download and install \u00ab\u00a0Visual Studio Build Tools 2022\u00a0\u00bb (free from Microsoft).<\/p>\n<p>Crucial: During installation, check the box for \u00ab\u00a0Desktop Development with C++\u00a0\u00bb. This installs the MSVC compiler needed to turn code into an application.<\/p>\n<h3>Miniconda (The Python Manager)<\/h3>\n<p>Install Miniconda for Windows (Python 3.x).<\/p>\n<p>Check the option \u00ab\u00a0Add Miniconda to my PATH environment variable\u00a0\u00bb (even if it&rsquo;s red-warned, it&rsquo;s simpler for automation here), or use the \u00ab\u00a0Anaconda Prompt\u00a0\u00bb to run the script.<\/p>\n<h3>Git<\/h3>\n<p>Install Git for Windows if not already done. Here&rsquo;s the link: https:\/\/git-scm.com\/download\/win<\/p>\n<p>Using Gemini to create a dedicated installer<br \/><img loading=\"lazy\" decoding=\"async\" width=\"696\" height=\"340\" class=\"wp-image-14780\" src=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-2.png\" srcset=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-2.png 696w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-2-300x147.png 300w\" sizes=\"auto, (max-width: 696px) 100vw, 696px\" \/><\/p>\n<p><img loading=\"lazy\" decoding=\"async\" width=\"668\" height=\"110\" class=\"wp-image-14781\" src=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-3.png\" srcset=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-3.png 668w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-3-300x49.png 300w\" sizes=\"auto, (max-width: 668px) 100vw, 668px\" \/><\/p>\n<p><img loading=\"lazy\" decoding=\"async\" width=\"662\" height=\"103\" class=\"wp-image-14782\" src=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-4.png\" srcset=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-4.png 662w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-4-300x47.png 300w\" sizes=\"auto, (max-width: 662px) 100vw, 662px\" \/><\/p>\n<p>A few attempts were necessary to have the final intall.bat script<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" width=\"709\" height=\"417\" class=\"wp-image-14783\" src=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-5.png\" srcset=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-5.png 709w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-5-300x176.png 300w\" sizes=\"auto, (max-width: 709px) 100vw, 709px\" \/><\/p>\n<h1>Launching the process<\/h1>\n<p>Use Anaconda Powershell to star the process<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" width=\"697\" height=\"114\" class=\"wp-image-14784\" src=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-6.png\" srcset=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-6.png 697w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-6-300x49.png 300w\" sizes=\"auto, (max-width: 697px) 100vw, 697px\" \/><\/p>\n<p><img loading=\"lazy\" decoding=\"async\" width=\"674\" height=\"89\" class=\"wp-image-14785\" src=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-7.png\" srcset=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-7.png 674w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-7-300x40.png 300w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-7-672x89.png 672w\" sizes=\"auto, (max-width: 674px) 100vw, 674px\" \/><\/p>\n<p>To work from a video file the command is<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" width=\"698\" height=\"125\" class=\"wp-image-14786\" src=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-8.png\" srcset=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-8.png 698w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-8-300x54.png 300w\" sizes=\"auto, (max-width: 698px) 100vw, 698px\" \/><em>ns-process-data video &#8211;data C:\\Splat\\test.mp4 &#8211;output-dir C:\\Splat\\my_project<\/em><\/p>\n<p>Create the 3D<\/p>\n<p><em>ns-train splatfacto &#8211;data C:\\Splat\\ my_project<\/em><\/p>\n<h2>Important verification (FFmpeg)<\/h2>\n<p><img loading=\"lazy\" decoding=\"async\" width=\"691\" height=\"107\" class=\"wp-image-14787\" src=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-9.png\" srcset=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-9.png 691w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-9-300x46.png 300w\" sizes=\"auto, (max-width: 691px) 100vw, 691px\" \/><\/p>\n<p><img loading=\"lazy\" decoding=\"async\" width=\"662\" height=\"111\" class=\"wp-image-14788\" src=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-10.png\" srcset=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-10.png 662w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-10-300x50.png 300w\" sizes=\"auto, (max-width: 662px) 100vw, 662px\" \/><\/p>\n<p><em>ffmpeg \u2013version<\/em><\/p>\n<p><em>conda install -c conda-forge ffmpeg \u2013y<\/em><\/p>\n<h2>Using Gemini to create a launcher<\/h2>\n<p>Gemini is used to create a launcher to replace the command line by buttons.<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" width=\"597\" height=\"831\" class=\"wp-image-14789\" src=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-11.png\" srcset=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-11.png 597w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-11-216x300.png 216w\" sizes=\"auto, (max-width: 597px) 100vw, 597px\" \/><\/p>\n<p>10 attempts have been necessary to have the final version. Make your own depending on your computer.<\/p>\n<h1>Step 2 : computing<\/h1>\n<p>First choose the photo set or video.<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" width=\"1348\" height=\"792\" class=\"wp-image-14790\" src=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-12.png\" srcset=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-12.png 1348w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-12-300x176.png 300w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-12-1024x602.png 1024w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-12-768x451.png 768w\" sizes=\"auto, (max-width: 1348px) 100vw, 1348px\" \/><\/p>\n<p>First is copies the photos or extract photos from the video<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" width=\"1094\" height=\"264\" class=\"wp-image-14791\" src=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-13.png\" srcset=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-13.png 1094w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-13-300x72.png 300w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-13-1024x247.png 1024w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-13-768x185.png 768w\" sizes=\"auto, (max-width: 1094px) 100vw, 1094px\" \/><\/p>\n<p><img loading=\"lazy\" decoding=\"async\" width=\"653\" height=\"210\" class=\"wp-image-14792\" src=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-14.png\" srcset=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-14.png 653w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-14-300x96.png 300w\" sizes=\"auto, (max-width: 653px) 100vw, 653px\" \/><\/p>\n<p>Then the COLMAP process starts, it can take a while<\/p>\n<p><strong>COLMAP<\/strong>\u00a0(CO<strong>LM<\/strong>A<strong>P<\/strong>) is a\u00a0<strong>general-purpose Structure-from-Motion (SfM)<\/strong>\u00a0and\u00a0<strong>Multi-View Stereo (MVS)<\/strong>\u00a0pipeline. It&rsquo;s the\u00a0<strong>de facto standard<\/strong>\u00a0for extracting 3D information from 2D images and is\u00a0<strong>crucial<\/strong>\u00a0for most NeRF and Gaussian Splatting workflows.<\/p>\n<p><strong>What Problem Does It Solve?<\/strong><\/p>\n<p>Given:\u00a0<strong>A set of unordered 2D images<\/strong>\u00a0of a scene<br \/>Produces:\u00a0<strong>Camera poses<\/strong>\u00a0+\u00a0<strong>3D point cloud<\/strong>\u00a0+\u00a0<strong>intrinsic parameters<\/strong><br \/>Essentially: It figures out\u00a0<strong>where each photo was taken from<\/strong>\u00a0and\u00a0<strong>what the 3D scene looks like<\/strong><\/p>\n<p><strong>The Three-Stage Pipeline<\/strong><\/p>\n<p><strong>Stage 1: Feature Extraction &amp; Matching<\/strong><\/p>\n<p><strong>Feature Detection<\/strong><\/p>\n<ul>\n<li>Extracts distinctive points (keypoints) from each image<\/li>\n<li>Uses algorithms like\u00a0<strong>SIFT<\/strong>\u00a0(Scale-Invariant Feature Transform) or\u00a0<strong>SuperPoint<\/strong><\/li>\n<li>Each keypoint has a descriptor (fingerprint) for matching<\/li>\n<\/ul>\n<p>Without COLMAP (or similar SfM), NeRF and Gaussian Splatting have\u00a0<strong>no idea<\/strong>:<\/p>\n<ul>\n<li>Where the camera was for each photo<\/li>\n<li>What the camera&rsquo;s lens properties were<\/li>\n<li>Initial 3D structure of the scene<\/li>\n<\/ul>\n<p><strong>COLMAP provides the foundational 3D understanding<\/strong>\u00a0that neural methods then refine and complete. It&rsquo;s the\u00a0<strong>bridge<\/strong>\u00a0between 2D images and 3D neural representations.<\/p>\n<p>The process is computationally expensive but necessary for high-quality results. Recent advances aim to reduce this dependency (like learning SfM end-to-end), but COLMAP remains the gold standard for photogrammetry-based 3D reconstruction.<\/p>\n<h1>Step 2 : Then the training starts<\/h1>\n<p>The web interface opens<\/p>\n<p>The interface allows to play with quality and duration<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" width=\"583\" height=\"153\" class=\"wp-image-14793\" src=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-15.png\" srcset=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-15.png 583w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-15-300x79.png 300w\" sizes=\"auto, (max-width: 583px) 100vw, 583px\" \/><\/p>\n<p><strong>Gaussian Splatting Training Process (Short Version)<\/strong><\/p>\n<p><strong>Training transforms a sparse 3D point cloud into millions of optimized Gaussian \u00ab\u00a0blobs\u00a0\u00bb:<\/strong><\/p>\n<ol>\n<li><strong>Start<\/strong>: Take COLMAP&rsquo;s sparse point cloud (camera poses + initial points)<\/li>\n<li><strong>Initialize<\/strong>: Convert each 3D point into a Gaussian with position, color, opacity, and initial size\/orientation<\/li>\n<li><strong>Iterative Optimization<\/strong>\u00a0(repeat 7,000-30,000 times):\n<ul>\n<li><strong>Project<\/strong>: \u00ab\u00a0Splat\u00a0\u00bb all Gaussians onto 2D image planes<\/li>\n<li><strong>Render<\/strong>: Alpha-blend them into synthetic images<\/li>\n<li><strong>Compare<\/strong>: Calculate loss vs. ground truth photos<\/li>\n<li><strong>Backpropagate<\/strong>: Update Gaussian parameters (position, color, opacity, shape)<\/li>\n<li><strong>Adapt<\/strong>: Clone Gaussians in detailed areas, prune transparent ones<\/li>\n<\/ul>\n<\/li>\n<li><strong>Result<\/strong>: A set of Gaussians that render photorealistic novel views at 100+ FPS<\/li>\n<\/ol>\n<p><strong>Key insight<\/strong>: Each Gaussian learns to represent a small patch of surface\/volume, optimizing position, shape, color, and transparency to match all training views simultaneously.<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" width=\"1092\" height=\"509\" class=\"wp-image-14794\" src=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-16.png\" srcset=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-16.png 1092w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-16-300x140.png 300w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-16-1024x477.png 1024w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-16-768x358.png 768w\" sizes=\"auto, (max-width: 1092px) 100vw, 1092px\" \/><\/p>\n<p>The web interface shows this :<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" width=\"1694\" height=\"1168\" class=\"wp-image-14795\" src=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-17.png\" srcset=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-17.png 1694w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-17-300x207.png 300w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-17-1024x706.png 1024w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-17-768x530.png 768w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-17-1536x1059.png 1536w\" sizes=\"auto, (max-width: 1694px) 100vw, 1694px\" \/><\/p>\n<p><img loading=\"lazy\" decoding=\"async\" width=\"1690\" height=\"1121\" class=\"wp-image-14796\" src=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-18.png\" srcset=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-18.png 1690w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-18-300x199.png 300w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-18-1024x679.png 1024w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-18-768x509.png 768w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-18-1536x1019.png 1536w\" sizes=\"auto, (max-width: 1690px) 100vw, 1690px\" \/><\/p>\n<h2>Exporting the PLY<\/h2>\n<p>The interface helps the export process<\/p>\n<p><em>ns-export gaussian-splat &#8211;load-config outputs\\unnamed\\splatfacto\\2026-01-04_132232\\config.yml &#8211;output-dir exports\/splat\/ <\/em><\/p>\n<p><img loading=\"lazy\" decoding=\"async\" width=\"588\" height=\"99\" class=\"wp-image-14797\" src=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-19.png\" srcset=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-19.png 588w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-19-300x51.png 300w\" sizes=\"auto, (max-width: 588px) 100vw, 588px\" \/><\/p>\n<p><img loading=\"lazy\" decoding=\"async\" width=\"615\" height=\"33\" class=\"wp-image-14798\" src=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-20.png\" srcset=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-20.png 615w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-20-300x16.png 300w\" sizes=\"auto, (max-width: 615px) 100vw, 615px\" \/><\/p>\n<h1>Editing the PLY file<\/h1>\n<p>Use Supersplat editor to clean the PLY file <a href=\"https:\/\/superspl.at\/editor\">https:\/\/superspl.at\/editor<\/a><\/p>\n<p><img loading=\"lazy\" decoding=\"async\" width=\"1691\" height=\"1164\" class=\"wp-image-14799\" src=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-21.png\" srcset=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-21.png 1691w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-21-300x207.png 300w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-21-1024x705.png 1024w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-21-768x529.png 768w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-21-1536x1057.png 1536w\" sizes=\"auto, (max-width: 1691px) 100vw, 1691px\" \/><\/p>\n<p>Here it\u2019s possible to rotate, reoriente the model.<\/p>\n<p>Also to render<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" width=\"1443\" height=\"1030\" class=\"wp-image-14800\" src=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/d-laurent-downloads-splat-image-2-png.png\" alt=\"D:\\LAURENT\\Downloads\\splat-image (2).png\" srcset=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/d-laurent-downloads-splat-image-2-png.png 1443w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/d-laurent-downloads-splat-image-2-png-300x214.png 300w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/d-laurent-downloads-splat-image-2-png-1024x731.png 1024w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/d-laurent-downloads-splat-image-2-png-768x548.png 768w\" sizes=\"auto, (max-width: 1443px) 100vw, 1443px\" \/><\/p>\n<p>After cleaning<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" width=\"1691\" height=\"1123\" class=\"wp-image-14801\" src=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-23.png\" srcset=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-23.png 1691w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-23-300x199.png 300w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-23-1024x680.png 1024w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-23-768x510.png 768w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-23-1536x1020.png 1536w\" sizes=\"auto, (max-width: 1691px) 100vw, 1691px\" \/><\/p>\n<p><img loading=\"lazy\" decoding=\"async\" width=\"1691\" height=\"1118\" class=\"wp-image-14802\" src=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-24.png\" srcset=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-24.png 1691w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-24-300x198.png 300w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-24-1024x677.png 1024w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-24-768x508.png 768w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-24-1536x1016.png 1536w\" sizes=\"auto, (max-width: 1691px) 100vw, 1691px\" \/> <img loading=\"lazy\" decoding=\"async\" width=\"1443\" height=\"1030\" class=\"wp-image-14803\" src=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/d-laurent-downloads-splat-image-4-png.png\" alt=\"D:\\LAURENT\\Downloads\\splat-image (4).png\" srcset=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/d-laurent-downloads-splat-image-4-png.png 1443w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/d-laurent-downloads-splat-image-4-png-300x214.png 300w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/d-laurent-downloads-splat-image-4-png-1024x731.png 1024w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/d-laurent-downloads-splat-image-4-png-768x548.png 768w\" sizes=\"auto, (max-width: 1443px) 100vw, 1443px\" \/><\/p>\n<p>Exporting the PLY<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" width=\"978\" height=\"661\" class=\"wp-image-14804\" src=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-26.png\" srcset=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-26.png 978w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-26-300x203.png 300w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-26-768x519.png 768w\" sizes=\"auto, (max-width: 978px) 100vw, 978px\" \/><\/p>\n<p>Visualising with https:\/\/urbandecoders-gaussiansplatviewer.netlify.app\/<\/p>\n<p>It\u2019s also possible from SuperSplat to create a HTML page to visualise<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" width=\"399\" height=\"339\" class=\"wp-image-14806\" src=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-28.png\" srcset=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-28.png 399w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-28-300x255.png 300w\" sizes=\"auto, (max-width: 399px) 100vw, 399px\" \/><\/p>\n<p><img loading=\"lazy\" decoding=\"async\" width=\"1687\" height=\"1166\" class=\"wp-image-14807\" src=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-29.png\" srcset=\"https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-29.png 1687w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-29-300x207.png 300w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-29-1024x708.png 1024w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-29-768x531.png 768w, https:\/\/www.keris-studio.fr\/blog\/wp-content\/word-image-14778-29-1536x1062.png 1536w\" sizes=\"auto, (max-width: 1687px) 100vw, 1687px\" \/><\/p>\n\n<div style=\"text-align: center\" class=\"align wp-block-tdvb-td-viewer\">\n\t<div class='tdvb3DViewerBlock'>\n\t\t<model-viewer\n\t\t\tsrc=''\n\t\t\talt=''\n\t\t\tcamera-controls\t\t\t\t\t\tloading='auto'\n\t\t\tauto-rotate\n\t\t\tstyle='width: 100%; height: 350px;'\n\t\t><\/model-viewer>\n\t<\/div>\n<\/div>","protected":false},"excerpt":{"rendered":"<p>Creating a Gaussian Splatting workflow Gaussian Splatting\u00a0is a novel 3D scene representation technique that&rsquo;s revolutionizing novel view synthesis (creating new camera viewpoints of 3D scenes). It emerged as a state-of-the-art alternative to Neural Radiance Fields (NeRFs) in 2023. Core Concept Instead of representing scenes with neural networks (like NeRFs) or explicit meshes, Gaussian Splatting represents &hellip; <a href=\"https:\/\/www.keris-studio.fr\/blog\/?p=14778\" class=\"more-link\">Continuer la lecture de <span class=\"screen-reader-text\">Gaussian Splatting: A High-Quality 3D Scene Representation<\/span>  <span class=\"meta-nav\">&rarr;<\/span><\/a><\/p>\n","protected":false},"author":2,"featured_media":14808,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[6,117,593,87,8,491],"tags":[244,546,623,419,624],"class_list":["post-14778","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-3d","category-archeologie","category-artificial","category-linkedin","category-methodologie","category-navale","tag-archeology","tag-artificial-intelligence","tag-gaussian-splatting","tag-ia","tag-nerf"],"_links":{"self":[{"href":"https:\/\/www.keris-studio.fr\/blog\/index.php?rest_route=\/wp\/v2\/posts\/14778","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.keris-studio.fr\/blog\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.keris-studio.fr\/blog\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.keris-studio.fr\/blog\/index.php?rest_route=\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.keris-studio.fr\/blog\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=14778"}],"version-history":[{"count":3,"href":"https:\/\/www.keris-studio.fr\/blog\/index.php?rest_route=\/wp\/v2\/posts\/14778\/revisions"}],"predecessor-version":[{"id":14929,"href":"https:\/\/www.keris-studio.fr\/blog\/index.php?rest_route=\/wp\/v2\/posts\/14778\/revisions\/14929"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.keris-studio.fr\/blog\/index.php?rest_route=\/wp\/v2\/media\/14808"}],"wp:attachment":[{"href":"https:\/\/www.keris-studio.fr\/blog\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=14778"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.keris-studio.fr\/blog\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=14778"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.keris-studio.fr\/blog\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=14778"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}