我們已經進入超視頻化時代，AI技術如何成爲視頻生產的利器？

{"type":"doc","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"視頻雲誕生於視頻化的大時代背景下，當前，全趨勢的視頻化、線上化，驅動着全新的超視頻化時代，而視頻雲所打造的音視頻數智化能力，在不斷延展新技術、締造新物種。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"對於競爭激烈、快速迭代的大視頻產業，視頻雲技術已經逐漸發展成大視頻產業的關鍵底座。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"7月10日，2021阿里雲視頻雲全景創新峯會暨全球視頻雲創新挑戰賽決賽頒獎典禮在北京舉辦。現場，阿里雲智能視頻雲負責人林昊表示，“融合了5G、雲、AI、音視頻，乃至虛擬技術，視頻雲是一個交叉學科領域，如何在這樣的融合基礎上重塑技術的新價值和商業的新方向，是視頻雲賽道關注的重點”。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/85\/93\/85023eb72825c38a15161d29e2d94793.png","alt":null,"title":"","style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":"center","origin":null},"content":[{"type":"text","text":"阿里雲智能視頻雲負責人林昊"}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"超視頻化時代"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"阿里雲智能視頻雲負責人林昊在演講中，聚焦視頻化在新內容和新交互的演進歷程，暢談場景的革新和創造，探索驅動體驗升級的關鍵技術。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"（以下爲林昊演講精華內容，有刪減，經編輯）"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"自今年開始，阿里雲特別強調雲上創新。因爲，很多企業已經走過了上雲的階段，很多企業的業務天然就在雲上誕生，所以現在最大的問題是，"},{"type":"text","marks":[{"type":"strong"}],"text":"在雲上誕生的業務怎麼繼續創新"},{"type":"text","text":"？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"阿里雲認爲，基於視頻可以非常好的往前推進創新，"},{"type":"text","marks":[{"type":"strong"}],"text":"視頻是全景式的創新"},{"type":"text","text":"。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"現在社會，各項數據表明，基於視頻做各種各樣的業務場景的趨勢持續上升。這意味着，與以前的視頻化相比，現在已經"},{"type":"text","marks":[{"type":"strong"},{"type":"strong"}],"text":"進入"},{"type":"text","text":"了"},{"type":"text","marks":[{"type":"strong"},{"type":"strong"}],"text":"超視頻化時代"},{"type":"text","text":"，內容"},{"type":"text","marks":[{"type":"strong"},{"type":"strong"}],"text":"更多"},{"type":"text","text":"地"},{"type":"text","marks":[{"type":"strong"},{"type":"strong"}],"text":"往視頻演進，交互的形式"},{"type":"text","text":"也"},{"type":"text","marks":[{"type":"strong"},{"type":"strong"}],"text":"更多"},{"type":"text","text":"了。此外，5G技術的發展，更好地解決延時及帶寬問題，帶來更好的超高清視頻體驗。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"林昊表示，雲和視頻的結合，體現在三個方面。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"第一在邊緣端做處理，進一步降低交互延時，提升很多場景下視頻的效果。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"此外，AI與視頻結合是現在的視頻化與過去相比最大的不同點，如在生產製作環節的智能翻譯、智能字幕等，AI可以輔助視頻製作。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"第三，混合現實是未來的發展趨勢。在內容上的演進和在交互上的演進是最強的訴求，而目前視頻是可以推進內容往前演進及交互演進，最成熟的一項技術。例如AR的發展能提供更加沉浸式的內容和交互形態。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"現在交互在往沉浸式方向不斷演進，未來人與人之間的交互可能不再是線下或者線上的視頻，交互可能會很難感受到虛擬的差別，"},{"type":"text","marks":[{"type":"strong"}],"text":"數字跟物理的環境"},{"type":"text","marks":[{"type":"strong"},{"type":"strong"},{"type":"strong"}],"text":"將"},{"type":"text","marks":[{"type":"strong"}],"text":"更加共存化"},{"type":"text","text":"。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"從技術層面看，內容和交互在融合式的演進。"},{"type":"text","text":" 因爲內容和交互在技術層面的很多東西很難完全分開，但對用戶來講，大多數情況下感受到的是，內容和交互兩者的結合體，因爲，以後內容會越來越帶交互性。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"視頻在技術領域的要求跟其他很多技術領域不同，"},{"type":"text","marks":[{"type":"strong"}],"text":"視頻雲涉及到非常多技術領域，"},{"type":"text","text":" 視頻是一個對AI、大數據、視頻本身的技術還有背後的分佈式系統都有全面要求的技術體系，不是某個單點能夠構成的技術。另外，視頻雲不僅僅是雲端的技術，在移動端、各種端設備上的整個音視頻的能力也都非常重要。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"目前我們可以看到，音視頻技術往前繼續演進，要做更好的內容形式，更好的交互形式。在現在的視頻整體技術上都要不斷的演進。其中最重要的一點是，以前的一張大面積的網絡主要服務於偏靜態和簡單的動態處理，現在大家都希望構建一張專門用來更好地，可以全球處理音視頻網絡的網，因爲這對於交互的延時及速度來說非常重要。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"此外，內容方向不論在朝清晰化、三維化還是全息化，更沉浸式的方向演進的時候，視頻在各塊技術領域都需要不斷的往前進步，否則很難落地，交互也同樣。"}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"AI技術驅動高清“視”界"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"達摩院資深算法專家謝宣松的分享中，他就達摩院在超高清的生產實踐展開介紹，通過AI關鍵技術和典型算法，深度剖析了AI技術如何驅動視聽升級。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"（以下爲謝宣松演講精華內容，有刪減，經編輯）"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"AI能在視頻中發揮什麼作用？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"達摩院資深算法專家謝宣松表示，AI發揮的作用主要有兩個："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"第一，基礎作用，即對於視頻或圖像的理解，包括分類、打標、檢測、分割等。第二，與生產類相關，如生產、編輯、加工、擦除、擦入等操作，還包括底層視覺相關的操作。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"那麼，AI如何在底層視覺上賦能視頻？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"對於視覺來說，體驗很重要。體驗又與分辨率、色彩、沉浸式等因素相關。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"2014年，國家發佈了4K戰略，7年間，在從4K往8K發展的過程中，內容永遠是滯後的，基礎設施反而是超前的。在內容生產過程中，技術能夠發揮很大的作用，比如要做視覺，最基本的要觀察細節，跟細節最相關的是分辨率，第二能不能流暢的看東西，流暢度不夠的信息怎麼能夠比較好的用技術的方法達成，第三關注色彩。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"從這三個層次來看採用的技術，追求細節可能會出現很多瑕疵，如很多學生常用高GAN系列技術，生產過程當中會經常出現不可控因素。視覺生產應該是可控的視覺內容生產，既能夠保證細節的還原又能夠保證瑕疵的控制，這是很核心的技術。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"另外，算法要控制超分辨率。算法的源頭是什麼？是數據。數據怎麼獲取？主要的辦法是用人工、高成本的方式，那麼，能否用技術的方式真正產生高仿真、高真實的數據對，這是很大的一個課題。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"最後，AI技術如何在實踐中使用，且平衡好效果和效率，這本身也是個問題。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"第一個要解決的是畫質。對於AI數據來講，數據規模龐大，在一些場景，大家感受到畫質比較差，在很多細節上損失掉了，可能有很多模糊、色彩不對的地方。所以，能不能設計一個方式，以自動的用算法獲取真實的數據，這是非常難的一個課題。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"接下來怎麼做增強？比如怎麼增強人像，我們現在比較重要的方法就是加入GAN系列技術進去，我們開源了GPEN的高清算法。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"那麼，怎麼解決流暢度的問題呢？在很多時候尤其在運動變化率非常大的時候經常出現一些瑕疵，所以，在不同的尺度，在不同的瑕疵時檢測到並彌補上，有很多事情可以做，除了細節增強，數據的還原、生成，瑕疵的檢測，差分的檢測，顏色，有非常多算法可以發揮作用，這一系列組合起來才能夠完成從低清到4K到8K，這本身也是個系統工程。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"AI驅動視頻往更高清的方向走的過程中，AI技術能不能自適應去做一些事情。自適應好像很簡單，但事實上在不同的場景，AI技術並沒有所謂的普適性的能力。因此，AI能在能夠自適應的針對不同場景採用最優質的算法，這是很重要的。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"視頻雲是一個基礎設施，我們有一個平臺，可以高效率、規模化的做各種各樣的AI視頻視覺增強的任務。"}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"解密優酷視頻智能生產"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"當時代對視頻的新文化消費不斷擴需，一方面，我們需要數字化的短視頻供應系統，一方面我們需要超高清視頻的生產能力，把大衆帶入數字化內容浪潮中、帶入到真正的8K時代。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"就此，阿里巴巴大文娛資深算法專家李靜從短視頻領域的內容生產困境出發，以概念級視頻解構賦能生產，分享Media AI平臺的技術能力及解密優酷短視頻智能生產。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"（以下爲李靜演講精華內容，有刪減，經編輯）"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"我要講的數字化一定會從視頻整個生命週期來講。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"作爲一個長視頻網站，從長視頻最初的拍攝到製作到完成，優酷有一個"},{"type":"text","marks":[{"type":"strong"}],"text":"內容評估的數字化系統。"},{"type":"text","marks":[{"type":"strong"},{"type":"strong"}],"text":"當一個長視頻拍攝出來之後，如何利用這種版權內容再進一步進行二次創作，這就是"},{"type":"text","marks":[{"type":"strong"}],"text":"創作的數字化"},{"type":"text","text":"。接下來在生成短視頻後，加一些特效，讓用戶看起來更好看，更有趣，這就是"},{"type":"text","marks":[{"type":"strong"}],"text":"特效的數字化。"},{"type":"text","marks":[{"type":"strong"},{"type":"strong"}],"text":"最後在終端上，用戶拿着手機或者通過平板或者通過電視大屏觀看視頻的時候如何進行互動，這是"},{"type":"text","marks":[{"type":"strong"}],"text":"互動的數字化。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"第一個內容是長視頻內容評估的數字化。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"在阿里大文娛，我們有一個北斗星團隊專注做長視頻的內容評估，它的核心點是希望通過後驗的數據來衡量整個視頻內容的質量，視頻生產的元素及外延信息。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"所有信息都基於我們ALP語音或者CV的能力進行解構。解構之後，通過這些信息去預測用戶的心理感受或者內容是不是受用戶的喜歡。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"所以從數據側我們獲得的先驗數據包括收視率，用戶的互動、評論等，但我們希望通過這些數據進一步挖掘用戶的心理狀態、生理狀態，以此推動我們的核心能力去做內容評估。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"具體應用的能力，一個是AI評估，一個是AI體檢。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"什麼是AI評估，比如一個視頻是否好，以前用人來審，但即便如此，去預測一個劇是否是爆款，還是很困難。所以，我們希望利用北斗星的系統，從最開始演員、供應商、IP等級、導演和編劇等信息去預估其水平。我們可以進一步對劇中的演員做進一步深度分析，如粉絲價值，口碑等。通過這些信息讓專家做輔助性決策，再利用AI技術幫助最終評估這部劇到底是什麼檔次。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"AI體檢，是指，當一個視頻拍完剪輯好後，我們希望用算法預測，哪些片斷是看點，哪些是高潮點，哪些是劇情拖沓的點.....找到用戶可能棄劇的風險點，並給出建設性的意見，幫助最終的剪輯進行優化。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如何進行二次創作？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"短視頻最近幾年非常火，去年短視頻用戶達到7億多，市場規模超過兩千億。在如此龐大的短視頻消費市場下，我們面臨了一些問題，"},{"type":"text","marks":[{"type":"strong"}],"text":"高質量的短視頻非常稀缺"},{"type":"text","text":"，大量的低質、粗製濫造的視頻存在市場上，所以我們想"},{"type":"text","marks":[{"type":"strong"}],"text":"用自動化生產的方式代替掉那些質量低劣的短視頻，讓智能創作接近於人創作的水平。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"我們研發出了"},{"type":"text","marks":[{"type":"strong"}],"text":"概念級的視頻的解構能力去賦能智能創作。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"每次提到視頻解構、CV能力，大家可能會自然聯想到一些標籤，如視頻場景中的人、美女、士兵等。在過去的CV領域，這些是特別客觀的標籤描述，但進行視頻創作時，這些東西並不是創作者們非常需要的，創作者們需要的元素和素材是能夠讓觀衆有深切感受的，如擁抱讓人感覺甜蜜，戰爭讓人覺得燃..... 所以我們重新定義了語義級的標籤，賦能智能創作。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"基於這種概念級的標籤解構能力之後，我們可以進行一系列的二次創作，如在某視頻片段中加入濃縮手段，減掉平淡的場景，對有意思的場景進行拼接、混剪等。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"我們過去這一年來所沉澱下來的AI技術，所有的視頻都可以實現自動化的生產。目前整個團隊的智能生產技術的產能，每天在萬條以上。每一條都要經過人工審覈，整個通過率爲90%，遠遠高於普通的優酷UP主上傳視頻的通過率。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"目前，這些二次創作的視頻每天VV量千萬級，虛擬UP主有百萬粉絲。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"下面講一下"},{"type":"text","marks":[{"type":"strong"}],"text":"劇情濃縮"},{"type":"text","text":"。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"現在大家碎片化的消費習慣，不太願意看長視頻了，我們針對這種情況採用了不同細粒度濃縮的手段，濃縮就是把關鍵劇情擇出來，在很短時間內讓你瞭解到故事的主線。比如，3分鐘看一部電影，15秒前情提要等。現在優酷每一個劇前面都有一個前情提要，那就是通過自動化生產出來的。我們具有風格化的解構能力，可以提取不同風格的前情提要，如對於女生做甜向的前情提要。對"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"針對解說類短視頻，我們重新定義了tts  video技術，把視頻進行解構化，同時把劇本進行解構化，兩者進行匹配，最後通過劇本生成視頻。解說的tts能力由達摩院提供。目前tts已經有10多種風格。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"接下來還有圖文轉視頻，配圖可以直接生成視頻，結合IP版權把圖片對應的視頻內容截出來。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"接下來講講特效的數字化。我們會針對於視頻中的人動作，動作的幅度，多人以及主角等進行自動化的加特效，這裏涉及到的CV技術包括動作檢測、動作幅度檢測、範圍檢測、明星識別、BGM、幅度檢測等。特效可以用於綜藝，也可以用於內生植入廣告，強化廣告注意點。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"另外在互動上，也有很多新玩法，第一個是視頻橫轉豎。例如在地鐵等擁擠的空間大家都是豎着看視頻，很少有橫着看的。基於這樣的需求，我們的技術可以講將橫視頻轉爲豎的。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"還有一個技術是自由視角視頻，這是國內也是業內第一個在C端用戶上可以做到體驗的自由視角的產品，利用這個技術，讓用戶多角度觀看他們想看的視頻內容。"}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"阿里雲天池數據集開源計劃發佈"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"超視頻化的時代，視頻雲的想象力不止步於商業場景，更多價值是普惠全民、創造的多元化的社會價值。就此，阿里雲聯手英特爾主辦、與優酷戰略技術合作的全球視頻雲創新挑戰賽於今年2月啓動，並在此次峯會上舉辦最終的頒獎典禮。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"據悉，本屆大賽吸引了全球4600支參賽戰隊。峯會上，阿里雲聯動技術方英特爾，對大賽 “算法“和”創新應用“兩大賽道，進行冠軍、亞軍、季軍、及最佳實踐獎的獎項頒佈。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"值得一提的是，通過與優酷平臺聯手，大賽提供了一個大規模高精度視頻分割數據集，供參賽選手訓練模型，並最終成功打磨成爲視頻分割領域的權威數據集"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"該數據集夯實的數據量級，共計有18萬幀以及高達30萬的最多視頻目標數據集，無論是標註精度還是內容廣度，均處於行業領先地位。此外，豐富的視頻來源覆蓋主流內容類型，高度貼合真實生產場景，囊括古裝、現代、街拍、街舞、運動等多場景，並對人物手持物、附屬物進行了細節標註，對於視頻產業具有高度實用意義。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"除此之外，天池平臺還整合全平臺資源，在本屆大會上推出了"},{"type":"text","marks":[{"type":"strong"}],"text":"數據集開源"},{"type":"text","text":"計劃，開放真實的業務場景和數據，希望打造專業的科研大數據平臺，應用在教育、社交、電商、醫療等衆多領域，併爲全球計算機視覺的人才培養做出貢獻。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.infoq.cn\/resource\/image\/96\/d7\/9682c83yy94834ecef1e3e28f97c4dd7.png","alt":null,"title":"","style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}}]}