{"id":88,"date":"2022-04-29T08:40:41","date_gmt":"2022-04-29T08:40:41","guid":{"rendered":"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/?page_id=88"},"modified":"2022-04-29T08:52:25","modified_gmt":"2022-04-29T08:52:25","slug":"research","status":"publish","type":"page","link":"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/research\/","title":{"rendered":"Alternative Approach"},"content":{"rendered":"\n<p>SliceNet<\/p>\n\n\n\n<p>An alternative approach to solving MVS using Sphere Sweep and Cost Volume Computation, would be to predict depth using a single panorama image view. Intuitively, this makes sense as in all cases, the model outputs a panorama depth map anyway.<\/p>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"362\" src=\"http:\/\/mscvprojects.ri.cmu.edu\/2022team6\/wp-content\/uploads\/sites\/61\/2022\/04\/image38-1024x362.png\" alt=\"\" class=\"wp-image-89\" srcset=\"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/wp-content\/uploads\/sites\/61\/2022\/04\/image38-1024x362.png 1024w, https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/wp-content\/uploads\/sites\/61\/2022\/04\/image38-300x106.png 300w, https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/wp-content\/uploads\/sites\/61\/2022\/04\/image38-768x272.png 768w, https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/wp-content\/uploads\/sites\/61\/2022\/04\/image38.png 1137w\" sizes=\"auto, (max-width: 706px) 89vw, (max-width: 767px) 82vw, 740px\" \/><figcaption>SliceNet<\/figcaption><\/figure>\n\n\n\n<p>SliceNet<sup><a href=\"https:\/\/openaccess.thecvf.com\/content\/CVPR2021\/papers\/Pintore_SliceNet_Deep_Dense_Depth_Estimation_From_a_Single_Indoor_Panorama_CVPR_2021_paper.pdf\" target=\"_blank\" rel=\"noreferrer noopener\">[1]<\/a><\/sup> estimates depth from a single input panorama image. <\/p>\n\n\n\n<ul class=\"wp-block-list\"><li>Panorama image is fed to a pretrained ResNet50 feature extractor.\u200b<\/li><li>The last 4 layer outputs are used to ensure that both, high level details and spatial context, are&nbsp;captured.\u200b<\/li><li>These outputs are passed through 3 asymmetric 1&#215;1 convolutional layers to reduce the channels&nbsp;and heights by a factor of 8.\u200b<\/li><li>The width component is then resized to 512 by interpolation and the reshaped components are&nbsp;concatenated to get 512 column slices of feature vectors of length 1024.\u200b<\/li><li>These slices sequentially represent the 3600&nbsp;view and so, are passed through a bi-directional&nbsp;LSTM setup.\u200b<\/li><li>The reshaped output is then&nbsp;upsampled&nbsp;to obtain the depth map.\u200b<\/li><\/ul>\n\n\n\n<p>The authors used an Adaptive Reverse Huber Loss<sup>[2]<\/sup> to train this network.<\/p>\n\n\n\n<figure class=\"wp-block-gallery has-nested-images columns-default is-cropped wp-block-gallery-1 is-layout-flex wp-block-gallery-is-layout-flex\">\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"397\" height=\"75\" data-id=\"90\" src=\"http:\/\/mscvprojects.ri.cmu.edu\/2022team6\/wp-content\/uploads\/sites\/61\/2022\/04\/image39.png\" alt=\"\" class=\"wp-image-90\" srcset=\"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/wp-content\/uploads\/sites\/61\/2022\/04\/image39.png 397w, https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/wp-content\/uploads\/sites\/61\/2022\/04\/image39-300x57.png 300w\" sizes=\"auto, (max-width: 397px) 100vw, 397px\" \/><figcaption>Adaptive Reverse Huber Loss<\/figcaption><\/figure>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"437\" height=\"118\" data-id=\"91\" src=\"http:\/\/mscvprojects.ri.cmu.edu\/2022team6\/wp-content\/uploads\/sites\/61\/2022\/04\/image40.png\" alt=\"\" class=\"wp-image-91\" srcset=\"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/wp-content\/uploads\/sites\/61\/2022\/04\/image40.png 437w, https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/wp-content\/uploads\/sites\/61\/2022\/04\/image40-300x81.png 300w\" sizes=\"auto, (max-width: 437px) 100vw, 437px\" \/><figcaption>Penalizing Gradients<\/figcaption><\/figure>\n<\/figure>\n\n\n\n<p>This loss is essentially a combination of L1 and L2 loss. However, just this alone wasn&#8217;t enough. As per studies<sup>[3]<\/sup>, CNNs tend to lose details during tasks such as depth estimation. Thus, the training signal also included loss terms penalizing the gradient along X and Y. These gradients were calculated using horizontal and vertical sobel filters<sup>[4]<\/sup>.<\/p>\n\n\n\n<p><\/p>\n","protected":false},"excerpt":{"rendered":"<p>SliceNet An alternative approach to solving MVS using Sphere Sweep and Cost Volume Computation, would be to predict depth using a single panorama image view. Intuitively, this makes sense as in all cases, the model outputs a panorama depth map anyway. SliceNet[1] estimates depth from a single input panorama image. Panorama image is fed to &hellip; <\/p>\n<p class=\"link-more\"><a href=\"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/research\/\" class=\"more-link\">Continue reading<span class=\"screen-reader-text\"> &#8220;Alternative Approach&#8221;<\/span><\/a><\/p>\n","protected":false},"author":120,"featured_media":0,"parent":0,"menu_order":0,"comment_status":"closed","ping_status":"closed","template":"","meta":{"footnotes":""},"class_list":["post-88","page","type-page","status-publish","hentry"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v28.1 - https:\/\/yoast.com\/product\/yoast-seo-wordpress\/ -->\n<title>Alternative Approach - Omnidirectional Multi-view Stereo on Edge Devices<\/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\/2022team6\/research\/\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Alternative Approach - Omnidirectional Multi-view Stereo on Edge Devices\" \/>\n<meta property=\"og:description\" content=\"SliceNet An alternative approach to solving MVS using Sphere Sweep and Cost Volume Computation, would be to predict depth using a single panorama image view. Intuitively, this makes sense as in all cases, the model outputs a panorama depth map anyway. SliceNet[1] estimates depth from a single input panorama image. Panorama image is fed to &hellip; Continue reading &quot;Alternative Approach&quot;\" \/>\n<meta property=\"og:url\" content=\"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/research\/\" \/>\n<meta property=\"og:site_name\" content=\"Omnidirectional Multi-view Stereo on Edge Devices\" \/>\n<meta property=\"article:modified_time\" content=\"2022-04-29T08:52:25+00:00\" \/>\n<meta property=\"og:image\" content=\"http:\/\/mscvprojects.ri.cmu.edu\/2022team6\/wp-content\/uploads\/sites\/61\/2022\/04\/image38-1024x362.png\" \/>\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=\"2 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2022team6\\\/research\\\/\",\"url\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2022team6\\\/research\\\/\",\"name\":\"Alternative Approach - Omnidirectional Multi-view Stereo on Edge Devices\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2022team6\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2022team6\\\/research\\\/#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2022team6\\\/research\\\/#primaryimage\"},\"thumbnailUrl\":\"http:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2022team6\\\/wp-content\\\/uploads\\\/sites\\\/61\\\/2022\\\/04\\\/image38-1024x362.png\",\"datePublished\":\"2022-04-29T08:40:41+00:00\",\"dateModified\":\"2022-04-29T08:52:25+00:00\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2022team6\\\/research\\\/#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2022team6\\\/research\\\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2022team6\\\/research\\\/#primaryimage\",\"url\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2022team6\\\/wp-content\\\/uploads\\\/sites\\\/61\\\/2022\\\/04\\\/image38.png\",\"contentUrl\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2022team6\\\/wp-content\\\/uploads\\\/sites\\\/61\\\/2022\\\/04\\\/image38.png\",\"width\":1137,\"height\":402},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2022team6\\\/research\\\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2022team6\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Alternative Approach\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2022team6\\\/#website\",\"url\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2022team6\\\/\",\"name\":\"Omnidirectional Multi-view Stereo on Edge Devices\",\"description\":\"\",\"publisher\":{\"@id\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2022team6\\\/#organization\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2022team6\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"},{\"@type\":\"Organization\",\"@id\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2022team6\\\/#organization\",\"name\":\"Omnidirectional Multi-view Stereo on Edge Devices\",\"url\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2022team6\\\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2022team6\\\/#\\\/schema\\\/logo\\\/image\\\/\",\"url\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2022team6\\\/wp-content\\\/uploads\\\/sites\\\/61\\\/2022\\\/04\\\/cropped-icon.png\",\"contentUrl\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2022team6\\\/wp-content\\\/uploads\\\/sites\\\/61\\\/2022\\\/04\\\/cropped-icon.png\",\"width\":250,\"height\":250,\"caption\":\"Omnidirectional Multi-view Stereo on Edge Devices\"},\"image\":{\"@id\":\"https:\\\/\\\/mscvprojects.ri.cmu.edu\\\/2022team6\\\/#\\\/schema\\\/logo\\\/image\\\/\"}}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Alternative Approach - Omnidirectional Multi-view Stereo on Edge Devices","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\/2022team6\/research\/","og_locale":"en_US","og_type":"article","og_title":"Alternative Approach - Omnidirectional Multi-view Stereo on Edge Devices","og_description":"SliceNet An alternative approach to solving MVS using Sphere Sweep and Cost Volume Computation, would be to predict depth using a single panorama image view. Intuitively, this makes sense as in all cases, the model outputs a panorama depth map anyway. SliceNet[1] estimates depth from a single input panorama image. Panorama image is fed to &hellip; Continue reading \"Alternative Approach\"","og_url":"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/research\/","og_site_name":"Omnidirectional Multi-view Stereo on Edge Devices","article_modified_time":"2022-04-29T08:52:25+00:00","og_image":[{"url":"http:\/\/mscvprojects.ri.cmu.edu\/2022team6\/wp-content\/uploads\/sites\/61\/2022\/04\/image38-1024x362.png","type":"","width":"","height":""}],"twitter_card":"summary_large_image","twitter_misc":{"Est. reading time":"2 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/research\/","url":"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/research\/","name":"Alternative Approach - Omnidirectional Multi-view Stereo on Edge Devices","isPartOf":{"@id":"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/#website"},"primaryImageOfPage":{"@id":"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/research\/#primaryimage"},"image":{"@id":"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/research\/#primaryimage"},"thumbnailUrl":"http:\/\/mscvprojects.ri.cmu.edu\/2022team6\/wp-content\/uploads\/sites\/61\/2022\/04\/image38-1024x362.png","datePublished":"2022-04-29T08:40:41+00:00","dateModified":"2022-04-29T08:52:25+00:00","breadcrumb":{"@id":"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/research\/#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/research\/"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/research\/#primaryimage","url":"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/wp-content\/uploads\/sites\/61\/2022\/04\/image38.png","contentUrl":"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/wp-content\/uploads\/sites\/61\/2022\/04\/image38.png","width":1137,"height":402},{"@type":"BreadcrumbList","@id":"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/research\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/"},{"@type":"ListItem","position":2,"name":"Alternative Approach"}]},{"@type":"WebSite","@id":"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/#website","url":"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/","name":"Omnidirectional Multi-view Stereo on Edge Devices","description":"","publisher":{"@id":"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/#organization"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"},{"@type":"Organization","@id":"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/#organization","name":"Omnidirectional Multi-view Stereo on Edge Devices","url":"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/","logo":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/#\/schema\/logo\/image\/","url":"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/wp-content\/uploads\/sites\/61\/2022\/04\/cropped-icon.png","contentUrl":"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/wp-content\/uploads\/sites\/61\/2022\/04\/cropped-icon.png","width":250,"height":250,"caption":"Omnidirectional Multi-view Stereo on Edge Devices"},"image":{"@id":"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/#\/schema\/logo\/image\/"}}]}},"_links":{"self":[{"href":"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/wp-json\/wp\/v2\/pages\/88","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/wp-json\/wp\/v2\/pages"}],"about":[{"href":"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/wp-json\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/wp-json\/wp\/v2\/users\/120"}],"replies":[{"embeddable":true,"href":"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/wp-json\/wp\/v2\/comments?post=88"}],"version-history":[{"count":5,"href":"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/wp-json\/wp\/v2\/pages\/88\/revisions"}],"predecessor-version":[{"id":106,"href":"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/wp-json\/wp\/v2\/pages\/88\/revisions\/106"}],"wp:attachment":[{"href":"https:\/\/mscvprojects.ri.cmu.edu\/2022team6\/wp-json\/wp\/v2\/media?parent=88"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}