{"id":635574,"date":"2021-08-20T08:33:39","date_gmt":"2021-08-20T15:33:39","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/?post_type=msr-project&#038;p=635574"},"modified":"2022-12-22T12:43:09","modified_gmt":"2022-12-22T20:43:09","slug":"watch-for","status":"publish","type":"msr-project","link":"https:\/\/www.microsoft.com\/en-us\/research\/project\/watch-for\/","title":{"rendered":"Watch For"},"content":{"rendered":"<section class=\"mb-3 moray-highlight\">\n\t<div class=\"card-img-overlay mx-lg-0\">\n\t\t<div class=\"card-background  has-background- card-background--full-bleed\">\n\t\t\t<img loading=\"lazy\" decoding=\"async\" width=\"1920\" height=\"720\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2020\/02\/WatchFor_AI_header_08_2021_1920x720.jpg\" class=\"attachment-full size-full\" alt=\"WatchFor video stream pattern\" style=\"\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2020\/02\/WatchFor_AI_header_08_2021_1920x720.jpg 1920w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2020\/02\/WatchFor_AI_header_08_2021_1920x720-300x113.jpg 300w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2020\/02\/WatchFor_AI_header_08_2021_1920x720-1024x384.jpg 1024w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2020\/02\/WatchFor_AI_header_08_2021_1920x720-768x288.jpg 768w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2020\/02\/WatchFor_AI_header_08_2021_1920x720-1536x576.jpg 1536w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2020\/02\/WatchFor_AI_header_08_2021_1920x720-1600x600.jpg 1600w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2020\/02\/WatchFor_AI_header_08_2021_1920x720-240x90.jpg 240w\" sizes=\"(max-width: 1920px) 100vw, 1920px\" \/>\t\t<\/div>\n\t\t<!-- Foreground -->\n\t\t<div class=\"card-foreground d-flex mt-md-n5 my-lg-5 px-g px-lg-0\">\n\t\t\t<!-- Container -->\n\t\t\t<div class=\"container d-flex mt-md-n5 my-lg-5 align-self-center\">\n\t\t\t\t<!-- Card wrapper -->\n\t\t\t\t<div class=\"w-100 w-lg-col-5\">\n\t\t\t\t\t<!-- Card -->\n\t\t\t\t\t<div class=\"card material-md-card py-5 px-md-5\">\n\t\t\t\t\t\t<div class=\"card-body \">\n\t\t\t\t\t\t\t\n\t\t\t\t\t\t\t\n\n<h1 id=\"watch-for\" class=\"h2\">Watch For<\/h1>\n\n\n\n<p>Analyze, identify, and surface the most interesting parts of your media content in real-time<\/p>\n\n\t\t\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t<\/div>\n\t\t<\/div>\n\t<\/div>\n<\/section>\n\n\n\n\n\n<p>Watch For is a large-scale, low-cost, highly programmable media AI platform from Microsoft Research that analyzes images, videos, live streams, and other media content in real-time. Our infrastructure is designed to address a broad range of verticals from digital safety to media analytics. The platform is currently used in production by several large organizations at Microsoft, processing over 400M+ minutes of video content and 4B+ frames every month with a reach of hundreds of millions of users.<\/p>\n\n\n\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-is-layout-1 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<figure class=\"wp-block-image size-full is-style-default\"><img loading=\"lazy\" decoding=\"async\" width=\"624\" height=\"351\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2020\/02\/WatchFor1.png\" alt=\"Fortnite game play screenshot with overlapping labels identifying what information Watch For is processing\" class=\"wp-image-767284\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2020\/02\/WatchFor1.png 624w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2020\/02\/WatchFor1-300x169.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2020\/02\/WatchFor1-343x193.png 343w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2020\/02\/WatchFor1-240x135.png 240w\" sizes=\"(max-width: 624px) 100vw, 624px\" \/><\/figure>\n<\/div>\n\n\n\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<figure class=\"wp-block-image size-full is-style-default\"><img loading=\"lazy\" decoding=\"async\" width=\"624\" height=\"351\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2021\/08\/WatchFor4.png\" alt=\"Fortnite game play screenshot with overlapping labels identifying what information Watch For is processing\" class=\"wp-image-767293\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2021\/08\/WatchFor4.png 624w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2021\/08\/WatchFor4-300x169.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2021\/08\/WatchFor4-343x193.png 343w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2021\/08\/WatchFor4-240x135.png 240w\" sizes=\"(max-width: 624px) 100vw, 624px\" \/><\/figure>\n<\/div>\n<\/div>\n\n\n\n<h2 id=\"watch-for-production-releases\">Watch For production releases<\/h2>\n\n\n\n<ul class=\"wp-block-list\"><li>Digital Safety solutions for several organizations including Xbox, Flipgrid, Bing, MSN, and LinkedIn<\/li><li>Powering <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.bing.com\/search?q=league%20of%20legends%20live%20streams&qs=n&form=QBRE&msbsrank=2_2__0&sp=-1&ghc=1&pq=league%20of%20legends%20live%20strea&sc=2-28&sk=&cvid=7A5EE8B918FD4785BC68D3B45C27EC5B\" target=\"_blank\" rel=\"noopener\">Bing\u2019s live stream search<\/a> serving hundreds of millions of user queries<\/li><li>Powering the AI experiences in <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/blogs.msn.com\/introducing-the-msn-esports-hub\/\" target=\"_blank\" rel=\"noopener\">MSN Esports Hub<\/a>, including Search, Spotlight and Highlights<\/li><li>Powering&nbsp;Mixer\u2019s&nbsp;<a href=\"https:\/\/www.microsoft.com\/en-us\/garage\/blog\/2018\/09\/watch-for-hackathon-2017-winner-powers-mixers-massively-successful-hypezone\/\" target=\"_blank\" rel=\"noopener\">HypeZone<\/a>, monitoring and analyzing live video streams on behalf of tens of millions of users and notifying them when specified events occur<\/li><\/ul>\n\n\n\n<h2 id=\"how-does-watch-for-work\">How does Watch For work?<\/h2>\n\n\n\n<p>Watch For runs analysis pipelines at scale and efficiently across large volumes of content.<\/p>\n\n\n\n<p>There is no silver bullet to achieving high efficiency. A combination of different techniques in different parts of the pipeline provide significant savings. The techniques in Watch For can be broadly classified into three buckets.<\/p>\n\n\n\n<ol class=\"wp-block-list\"><li><strong>End-to-end resource utilization optimizations:<\/strong> Watch For optimizes resources at cluster-level, node-level, and process-level. At the cluster-level, Watch For efficiently manages both spot and dedicated instances and orchestrates between them to process content at low cost and low latency. At the node-level, Watch For makes sure network and CPUs are effectively utilized. The system\u2019s workload is a combination of network and processing, and it applies a few techniques to keep utilization high.<\/li><li><strong>ML optimizations:<\/strong> Watch For does ML optimizations such as batching, model cascades, and bit-width optimizations. Watch For team has been working closely with OctoML and piloting Apache TVM as a model runtime to achieve high inference efficiency.<\/li><li><strong>Efficient programming templates:<\/strong> Watch For exposes efficient programming templates for various content types and pipelines written using those templates are efficiently executed at large scale. Watch For also exposes as many knobs to the developer as possible so that optimizations such as adaptive sampling and deduping can be easily implemented.<\/li><\/ol>\n\n\n","protected":false},"excerpt":{"rendered":"<p>Transitioned | Watch For is a large-scale, low-cost, highly programmable media analysis platform built on Azure.<\/p>\n","protected":false},"featured_media":767359,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"_classifai_error":"","footnotes":""},"research-area":[13556],"msr-locale":[268875],"msr-impact-theme":[],"msr-pillar":[],"class_list":["post-635574","msr-project","type-msr-project","status-publish","has-post-thumbnail","hentry","msr-research-area-artificial-intelligence","msr-locale-en_us","msr-archive-status-complete"],"msr_project_start":"","related-publications":[563634,767323,767335,367685,496640],"related-downloads":[],"related-videos":[],"related-groups":[],"related-events":[],"related-opportunities":[],"related-posts":[],"related-articles":[],"tab-content":[],"slides":[],"related-researchers":[{"type":"user_nicename","display_name":"Matthai Philipose","user_id":32834,"people_section":"Section name 0","alias":"matthaip"},{"type":"user_nicename","display_name":"Peter Bod\u00edk","user_id":33239,"people_section":"Section name 0","alias":"peterb"},{"type":"user_nicename","display_name":"Lenin Ravindranath Sivalingam","user_id":32645,"people_section":"Section name 0","alias":"lenin"},{"type":"guest","display_name":"Chris Linseman","user_id":632463,"people_section":"Section name 0","alias":""},{"type":"user_nicename","display_name":"Loc Huynh","user_id":38904,"people_section":"Section name 0","alias":"lohuynh"},{"type":"user_nicename","display_name":"Weishung Liu","user_id":39805,"people_section":"Section name 0","alias":"weisliu"},{"type":"guest","display_name":"Sahil Desai","user_id":727525,"people_section":"Section name 0","alias":""},{"type":"guest","display_name":"Mina Kim","user_id":727528,"people_section":"Section name 0","alias":""},{"type":"guest","display_name":"Disa Mhembere","user_id":746926,"people_section":"Section name 0","alias":""},{"type":"guest","display_name":"Saewon Kwak","user_id":763327,"people_section":"Section name 0","alias":""},{"type":"guest","display_name":"Casey Zhang","user_id":763324,"people_section":"Section name 0","alias":""}],"msr_research_lab":[199565],"msr_impact_theme":[],"_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-project\/635574"}],"collection":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-project"}],"about":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-project"}],"version-history":[{"count":31,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-project\/635574\/revisions"}],"predecessor-version":[{"id":962961,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-project\/635574\/revisions\/962961"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media\/767359"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=635574"}],"wp:term":[{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=635574"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=635574"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=635574"},{"taxonomy":"msr-pillar","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-pillar?post=635574"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}