{"id":13,"date":"2025-05-08T03:12:45","date_gmt":"2025-05-08T03:12:45","guid":{"rendered":"https:\/\/mscvprojects.ri.cmu.edu\/2025team4\/?page_id=13"},"modified":"2025-05-09T18:00:27","modified_gmt":"2025-05-09T18:00:27","slug":"pipeline-2","status":"publish","type":"page","link":"https:\/\/mscvprojects.ri.cmu.edu\/2025team4\/pipeline-2\/","title":{"rendered":"Methodology"},"content":{"rendered":"\n<p><img decoding=\"async\" width=\"1527px;\" height=\"1306px;\" src=\"https:\/\/lh7-rt.googleusercontent.com\/slidesz\/AGV_vUecA_CUcbEeGiQQz0nCAzozfNBYr2BxIp8FmERxeAZi_UUjAGWCDmb8CnqPB6GiHS0p7zh1UYHoRpfp-J4LcVqpEhQbBxSYPhQxbmlvthZ7PgcTLBcKfEZj9K7SicZRjHgHBKYzOb45IEfCqtfZkg=s2048?key=nYJ-3L-7P7kP051xKoAdjm9F\"><\/p>\n\n\n\n<h1 class=\"wp-block-heading\">Data Collection Format: 360\u00b0 Cameras<\/h1>\n\n\n\n<p>To support high-fidelity reconstruction and robust SLAM, we use a <strong>monocular 360\u00b0 camera<\/strong> for outdoor data collection. Unlike traditional narrow field-of-view cameras (Handheld devices) which require careful path planning to ensure sufficient coverage, 360\u00b0 cameras capture the <strong>entire surrounding scene in every frame<\/strong>. This dramatically reduces blind spots and eliminates the need for complex multi-pass recordings, making them ideal for <strong>unstructured outdoor environments<\/strong>.<\/p>\n\n\n\n<p>In the context of <strong>Gaussian Splatting<\/strong>, where reconstruction quality depends heavily on <strong>view diversity and scene coverage<\/strong>, 360\u00b0 imagery offers <strong>dense multi-view supervision<\/strong> from a single trajectory. For <strong>SLAM<\/strong>, the panoramic input significantly improves <strong>feature persistence<\/strong> and <strong>loop closure detection<\/strong>, especially in textureless or repetitive regions. The result is a more efficient and robust pipeline, with fewer artifacts due to occlusions, missing views, or tracking drift.<\/p>\n\n\n\n<p>As part of our pipeline, users are free to <strong>walk naturally<\/strong> through a scene without following strict paths or capture protocols. Running SLAM directly on 360\u00b0 video enables reliable feature tracking even in narrow corridors or low-texture regions, resulting in improved robustness and trajectory accuracy.<\/p>\n\n\n\n<p>Since the current implementation runs offline, our pipeline supports <strong>any 360\u00b0 video<\/strong> (equirectangular or dual-fisheye formats), making it compatible with a wide range of commercially available cameras. Furthermore, handheld devices with pinhole cameras can be used to record and bolster under-constrained regions. <\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\" \/>\n\n\n\n<h1 class=\"wp-block-heading\">1. Preprocessing: SLAM<\/h1>\n\n\n\n<p>A core requirement for Gaussian Splatting is access to <strong>precise camera poses<\/strong> for each image. We begin with a <strong>SLAM-based preprocessing<\/strong> stage that outputs both trajectory and geometric structure.<\/p>\n\n\n\n<p><strong>Input:<\/strong><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>360\u00b0 monocular video (equirectangular or dual-fisheye)<\/li>\n<\/ul>\n\n\n\n<p><strong>Output:<\/strong><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Sparse point cloud from feature-based SLAM<\/li>\n\n\n\n<li>Dense point cloud reconstructed via <strong>PatchMatch Stereo<\/strong><\/li>\n\n\n\n<li>Estimated camera poses for all extracted frames<\/li>\n\n\n\n<li>6-perspective <strong>cubemap faces<\/strong> per 360\u00b0 frame<\/li>\n<\/ul>\n\n\n\n<p>Details of this stage are provided in the <strong>OpenVSLAM<\/strong> section, which outlines the mapping, tracking, and optimization procedures used.<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\" \/>\n\n\n\n<h1 class=\"wp-block-heading\">2. Postprocessing: Gaussian Splatting<\/h1>\n\n\n\n<p>After preprocessing, the cubemap views are passed into our <strong>Gaussian Splatting pipeline<\/strong> for 3D scene reconstruction. Outdoor data, captured under natural lighting, often contains significant <strong>photometric variation<\/strong>, which poses challenges for models trained with photometric consistency losses.<\/p>\n\n\n\n<p>To address this, we adopt a <strong>lighting-aware variant<\/strong> of Gaussian Splatting that includes:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Per-Gaussian neural color features<\/strong><\/li>\n\n\n\n<li><strong>Per-image appearance embeddings<\/strong><\/li>\n\n\n\n<li>A <strong>spherical harmonics-based background model<\/strong> to account for global lighting variation and sky\/background consistency<\/li>\n<\/ul>\n\n\n\n<p>This enhanced formulation improves robustness to environmental changes and enables <strong>photorealistic reconstructions<\/strong> from real-world, unconstrained imagery.<\/p>\n\n\n\n<p>Additional architectural details and training specifics are available in the <strong>&#8220;GS in the Wild&#8221;<\/strong> section.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">References<\/h2>\n\n\n\n<p>Hartmut Surmann, Marc Thurow, and Dominik Slomma. <em>PatchMatch-Stereo-Panorama: A Fast Dense Reconstruction from 360\u00b0 Video Images.<\/em> arXiv preprint arXiv:2211.16266, 2022. Available at: <a class=\"\" href=\"https:\/\/arxiv.org\/abs\/2211.16266\">https:\/\/arxiv.org\/abs\/2211.16266<\/a><\/p>\n\n\n\n<p>Congrong Xu, Justin Kerr, and Angjoo Kanazawa. <em>Splatfacto-W: A Nerfstudio Implementation of Gaussian Splatting for Unconstrained Photo Collections.<\/em> arXiv preprint arXiv:2407.12306, 2024. Available at: <a class=\"\" href=\"https:\/\/arxiv.org\/abs\/2407.12306\">https:\/\/arxiv.org\/abs\/2407.12306<\/a><\/p>\n\n\n\n<p><\/p>\n","protected":false},"excerpt":{"rendered":"<p>Data Collection Format: 360\u00b0 Cameras To support high-fidelity reconstruction and robust SLAM, we use a monocular 360\u00b0 camera for outdoor data collection. Unlike traditional narrow field-of-view cameras (Handheld devices) which require careful path planning to ensure sufficient coverage, 360\u00b0 cameras capture the entire surrounding scene in every frame. This dramatically reduces blind spots and eliminates &hellip; <\/p>\n<p class=\"link-more\"><a href=\"https:\/\/mscvprojects.ri.cmu.edu\/2025team4\/pipeline-2\/\" class=\"more-link\">Continue reading<span class=\"screen-reader-text\"> &#8220;Methodology&#8221;<\/span><\/a><\/p>\n","protected":false},"author":233,"featured_media":0,"parent":0,"menu_order":0,"comment_status":"closed","ping_status":"closed","template":"","meta":{"footnotes":""},"class_list":["post-13","page","type-page","status-publish","hentry"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v27.4 - https:\/\/yoast.com\/product\/yoast-seo-wordpress\/ -->\n<title>Methodology - Adaptive Data Collection For High Fidelity Gaussian Splatting<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/mscvprojects.ri.cmu.edu\/2025team4\/pipeline-2\/\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Methodology - Adaptive Data Collection For High Fidelity Gaussian Splatting\" \/>\n<meta property=\"og:description\" content=\"Data Collection Format: 360\u00b0 Cameras To support high-fidelity reconstruction and robust SLAM, we use a monocular 360\u00b0 camera for outdoor data collection. Unlike traditional narrow field-of-view cameras (Handheld devices) which require careful path planning to ensure sufficient coverage, 360\u00b0 cameras capture the entire surrounding scene in every frame. This dramatically reduces blind spots and eliminates &hellip; Continue reading &quot;Methodology&quot;\" \/>\n<meta property=\"og:url\" content=\"https:\/\/mscvprojects.ri.cmu.edu\/2025team4\/pipeline-2\/\" \/>\n<meta property=\"og:site_name\" content=\"Adaptive Data Collection For High Fidelity Gaussian Splatting\" \/>\n<meta property=\"article:modified_time\" content=\"2025-05-09T18:00:27+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/lh7-rt.googleusercontent.com\/slidesz\/AGV_vUecA_CUcbEeGiQQz0nCAzozfNBYr2BxIp8FmERxeAZi_UUjAGWCDmb8CnqPB6GiHS0p7zh1UYHoRpfp-J4LcVqpEhQbBxSYPhQxbmlvthZ7PgcTLBcKfEZj9K7SicZRjHgHBKYzOb45IEfCqtfZkg=s2048?key=nYJ-3L-7P7kP051xKoAdjm9F\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data1\" content=\"3 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2025team4\\\/pipeline-2\\\/\",\"url\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2025team4\\\/pipeline-2\\\/\",\"name\":\"Methodology - Adaptive Data Collection For High Fidelity Gaussian Splatting\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2025team4\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2025team4\\\/pipeline-2\\\/#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2025team4\\\/pipeline-2\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/lh7-rt.googleusercontent.com\\\/slidesz\\\/AGV_vUecA_CUcbEeGiQQz0nCAzozfNBYr2BxIp8FmERxeAZi_UUjAGWCDmb8CnqPB6GiHS0p7zh1UYHoRpfp-J4LcVqpEhQbBxSYPhQxbmlvthZ7PgcTLBcKfEZj9K7SicZRjHgHBKYzOb45IEfCqtfZkg=s2048?key=nYJ-3L-7P7kP051xKoAdjm9F\",\"datePublished\":\"2025-05-08T03:12:45+00:00\",\"dateModified\":\"2025-05-09T18:00:27+00:00\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2025team4\\\/pipeline-2\\\/#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2025team4\\\/pipeline-2\\\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2025team4\\\/pipeline-2\\\/#primaryimage\",\"url\":\"https:\\\/\\\/lh7-rt.googleusercontent.com\\\/slidesz\\\/AGV_vUecA_CUcbEeGiQQz0nCAzozfNBYr2BxIp8FmERxeAZi_UUjAGWCDmb8CnqPB6GiHS0p7zh1UYHoRpfp-J4LcVqpEhQbBxSYPhQxbmlvthZ7PgcTLBcKfEZj9K7SicZRjHgHBKYzOb45IEfCqtfZkg=s2048?key=nYJ-3L-7P7kP051xKoAdjm9F\",\"contentUrl\":\"https:\\\/\\\/lh7-rt.googleusercontent.com\\\/slidesz\\\/AGV_vUecA_CUcbEeGiQQz0nCAzozfNBYr2BxIp8FmERxeAZi_UUjAGWCDmb8CnqPB6GiHS0p7zh1UYHoRpfp-J4LcVqpEhQbBxSYPhQxbmlvthZ7PgcTLBcKfEZj9K7SicZRjHgHBKYzOb45IEfCqtfZkg=s2048?key=nYJ-3L-7P7kP051xKoAdjm9F\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2025team4\\\/pipeline-2\\\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2025team4\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Methodology\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2025team4\\\/#website\",\"url\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2025team4\\\/\",\"name\":\"Adaptive Data Collection For High Fidelity Gaussian Splatting\",\"description\":\"\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2025team4\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Methodology - Adaptive Data Collection For High Fidelity Gaussian Splatting","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/mscvprojects.ri.cmu.edu\/2025team4\/pipeline-2\/","og_locale":"en_US","og_type":"article","og_title":"Methodology - Adaptive Data Collection For High Fidelity Gaussian Splatting","og_description":"Data Collection Format: 360\u00b0 Cameras To support high-fidelity reconstruction and robust SLAM, we use a monocular 360\u00b0 camera for outdoor data collection. Unlike traditional narrow field-of-view cameras (Handheld devices) which require careful path planning to ensure sufficient coverage, 360\u00b0 cameras capture the entire surrounding scene in every frame. This dramatically reduces blind spots and eliminates &hellip; Continue reading \"Methodology\"","og_url":"https:\/\/mscvprojects.ri.cmu.edu\/2025team4\/pipeline-2\/","og_site_name":"Adaptive Data Collection For High Fidelity Gaussian Splatting","article_modified_time":"2025-05-09T18:00:27+00:00","og_image":[{"url":"https:\/\/lh7-rt.googleusercontent.com\/slidesz\/AGV_vUecA_CUcbEeGiQQz0nCAzozfNBYr2BxIp8FmERxeAZi_UUjAGWCDmb8CnqPB6GiHS0p7zh1UYHoRpfp-J4LcVqpEhQbBxSYPhQxbmlvthZ7PgcTLBcKfEZj9K7SicZRjHgHBKYzOb45IEfCqtfZkg=s2048?key=nYJ-3L-7P7kP051xKoAdjm9F","type":"","width":"","height":""}],"twitter_card":"summary_large_image","twitter_misc":{"Est. reading time":"3 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/mscvprojects.ri.cmu.edu\/2025team4\/pipeline-2\/","url":"https:\/\/mscvprojects.ri.cmu.edu\/2025team4\/pipeline-2\/","name":"Methodology - Adaptive Data Collection For High Fidelity Gaussian Splatting","isPartOf":{"@id":"https:\/\/mscvprojects.ri.cmu.edu\/2025team4\/#website"},"primaryImageOfPage":{"@id":"https:\/\/mscvprojects.ri.cmu.edu\/2025team4\/pipeline-2\/#primaryimage"},"image":{"@id":"https:\/\/mscvprojects.ri.cmu.edu\/2025team4\/pipeline-2\/#primaryimage"},"thumbnailUrl":"https:\/\/lh7-rt.googleusercontent.com\/slidesz\/AGV_vUecA_CUcbEeGiQQz0nCAzozfNBYr2BxIp8FmERxeAZi_UUjAGWCDmb8CnqPB6GiHS0p7zh1UYHoRpfp-J4LcVqpEhQbBxSYPhQxbmlvthZ7PgcTLBcKfEZj9K7SicZRjHgHBKYzOb45IEfCqtfZkg=s2048?key=nYJ-3L-7P7kP051xKoAdjm9F","datePublished":"2025-05-08T03:12:45+00:00","dateModified":"2025-05-09T18:00:27+00:00","breadcrumb":{"@id":"https:\/\/mscvprojects.ri.cmu.edu\/2025team4\/pipeline-2\/#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/mscvprojects.ri.cmu.edu\/2025team4\/pipeline-2\/"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/mscvprojects.ri.cmu.edu\/2025team4\/pipeline-2\/#primaryimage","url":"https:\/\/lh7-rt.googleusercontent.com\/slidesz\/AGV_vUecA_CUcbEeGiQQz0nCAzozfNBYr2BxIp8FmERxeAZi_UUjAGWCDmb8CnqPB6GiHS0p7zh1UYHoRpfp-J4LcVqpEhQbBxSYPhQxbmlvthZ7PgcTLBcKfEZj9K7SicZRjHgHBKYzOb45IEfCqtfZkg=s2048?key=nYJ-3L-7P7kP051xKoAdjm9F","contentUrl":"https:\/\/lh7-rt.googleusercontent.com\/slidesz\/AGV_vUecA_CUcbEeGiQQz0nCAzozfNBYr2BxIp8FmERxeAZi_UUjAGWCDmb8CnqPB6GiHS0p7zh1UYHoRpfp-J4LcVqpEhQbBxSYPhQxbmlvthZ7PgcTLBcKfEZj9K7SicZRjHgHBKYzOb45IEfCqtfZkg=s2048?key=nYJ-3L-7P7kP051xKoAdjm9F"},{"@type":"BreadcrumbList","@id":"https:\/\/mscvprojects.ri.cmu.edu\/2025team4\/pipeline-2\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/mscvprojects.ri.cmu.edu\/2025team4\/"},{"@type":"ListItem","position":2,"name":"Methodology"}]},{"@type":"WebSite","@id":"https:\/\/mscvprojects.ri.cmu.edu\/2025team4\/#website","url":"https:\/\/mscvprojects.ri.cmu.edu\/2025team4\/","name":"Adaptive Data Collection For High Fidelity Gaussian Splatting","description":"","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/mscvprojects.ri.cmu.edu\/2025team4\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"}]}},"_links":{"self":[{"href":"https:\/\/mscvprojects.ri.cmu.edu\/2025team4\/wp-json\/wp\/v2\/pages\/13","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/mscvprojects.ri.cmu.edu\/2025team4\/wp-json\/wp\/v2\/pages"}],"about":[{"href":"https:\/\/mscvprojects.ri.cmu.edu\/2025team4\/wp-json\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"https:\/\/mscvprojects.ri.cmu.edu\/2025team4\/wp-json\/wp\/v2\/users\/233"}],"replies":[{"embeddable":true,"href":"https:\/\/mscvprojects.ri.cmu.edu\/2025team4\/wp-json\/wp\/v2\/comments?post=13"}],"version-history":[{"count":8,"href":"https:\/\/mscvprojects.ri.cmu.edu\/2025team4\/wp-json\/wp\/v2\/pages\/13\/revisions"}],"predecessor-version":[{"id":142,"href":"https:\/\/mscvprojects.ri.cmu.edu\/2025team4\/wp-json\/wp\/v2\/pages\/13\/revisions\/142"}],"wp:attachment":[{"href":"https:\/\/mscvprojects.ri.cmu.edu\/2025team4\/wp-json\/wp\/v2\/media?parent=13"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}