2020年，那些令人印象深刻的AI論文

{"type":"doc","content":[{"type":"blockquote","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"儘管今年這世界發生很多事，但我們仍然有機會看到那麼多豐碩的研究成果，特別是在人工智能領域。同時，Al 偏見和 Al 倫理也開始逐漸進入大家的視野，引起大家的普遍重視。人工智能在不斷髮展，我們對人類大腦以及它與人工智能的聯繫的理解在不斷深入，在不久的將來勢必將出現了不起的應用。本文將爲你介紹本年度那些不容錯過的最有意思的研究論文，並附上了論文鏈接和相關代碼地址。"}]},{"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":"如果本文漏掉了那些重要的論文，歡迎在評價區留言告訴我和其他的讀者小夥伴！"}]},{"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":"文末列出了引用的所有論文"}]},{"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 GitHub 查閱完整列表："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020"}]}]},{"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":"觀看 15 分鐘時長的 2020 年度完整回放："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/youtu.be\/DHBclF-8KwE","title":"","type":null},"content":[{"type":"text","text":"https:\/\/youtu.be\/DHBclF-8KwE"}]}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"1、YOLOv4: 目標檢測的最佳速度和精度 [1]"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/bb\/bb0b96d38a377c33540ab0cfc06dd443.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"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":"Alexey Bochkovsky 等人於 2020 年 4 月在論文“YOLOv4: 目標檢測的最佳速度和精度”中介紹它的第 4 個版本。該算法的主要目標是製作一個高精度、高質量的超高速目標檢測器。"}]},{"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":"論文鏈接："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#1","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#1"}]}]},{"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":"代碼地址："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/AlexeyAB\/darknet","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/AlexeyAB\/darknet"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"2、DeepFace rawing：依據草圖深度生成人臉圖像 [2]"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/fb\/fb63054083be1cd5cc913eb8b67150f1.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"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":"你現在可以使用這種新的圖像到圖像的轉換技術，依據粗糙甚至不完整的草圖生成高質量的人臉圖像，而你本人無需繪圖技巧！如果你的畫技和我一樣差，你甚至可以調整眼睛、嘴巴和鼻子對最終圖像的影響。讓我們看看它是否真的有效，以及他們是如何做到的。"}]},{"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":"論文鏈接："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#2","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#2"}]}]},{"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":"代碼地址："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/IGLICT\/DeepFaceDrawing-Jittor","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/IGLICT\/DeepFaceDrawing-Jittor"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"3、英偉達研究人員用人工智能重新制作了《喫豆人》[3]"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/64\/64834051e5b710ba06b08e3228499b73.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"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":"40 年前，《喫豆人》首次登陸日本的街機平臺，併成爲全球巨星，現在英偉達研究人員用人工智能重新制作了它。（"},{"type":"link","attrs":{"href":"https:\/\/blogs.nvidia.com\/blog\/2020\/05\/22\/gamegan-research-pacman-anniversary\/%EF%BC%89","title":"","type":null},"content":[{"type":"text","text":"https:\/\/blogs.nvidia.com\/blog\/2020\/05\/22\/gamegan-research-pacman-anniversary\/）"}]}]},{"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":"論文鏈接："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#3","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#3"}]}]},{"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":"代碼地址："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/nv-tlabs\/GameGAN_code","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/nv-tlabs\/GameGAN_code"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"4、PULSE: 通過衍生模型的潛在空間探索進行自我監督的照片採樣提升 [4]"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/80\/80703c7f421b21caecf6e64aab0eaf3f.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"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":"這個新算法將模糊的圖像轉換成高分辨率的圖像! 它可以把超低分辨率的 16x16 圖像轉換成 1080p 高清晰的人臉，人工智能讓模糊的臉看起來清晰 60 倍！如若不信，就可以像我一樣親自試一試，只需不到一分鐘！但首先，讓我們看看他們是怎麼做到的。"}]},{"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":"論文鏈接："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#4","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#4"}]}]},{"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":"代碼地址："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/adamian98\/pulse","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/adamian98\/pulse"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"5、編程語言的無監督翻譯 [5]"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/67\/6794051a7c0340c7ef52f11170432f33.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"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":"這種新模型無需任何監督即可將代碼從一種編程語言轉換成另一種編程語言！它可以將一個 Python 函數轉換成 c++ 函數，反之亦然，而不需要任何先前的例子！它理解每種語言的語法，因此可以推廣到任何編程語言！我們來看看他們是怎麼做到的。"}]},{"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":"論文鏈接："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#5","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#5"}]}]},{"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":"代碼地址："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/facebookresearch\/TransCoder?utm_source=catalyzex.com","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/facebookresearch\/TransCoder?utm_source=catalyzex.com"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"6、用於高分辨率三維人體重建的多層次像素對齊隱式函數 [6]"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/56\/56b103811edb64f0c391665ff67cf99f.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"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":"這個人工智能通過 2 維圖像重新生成 3 維高分辨率的人體！你只需要一張圖片，哪怕是背面的，它就能生成一個三維的你，看起來一模一樣！"}]},{"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":"論文鏈接："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#6","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#6"}]}]},{"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":"代碼地址："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/facebookresearch\/pifuhd","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/facebookresearch\/pifuhd"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"7、高分辨率換臉技術 [7]"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/3e\/3ece923f2bb23489373c798a938bf585.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"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":"迪士尼的研究人員在同名論文中開發了一種新的高分辨率換臉算法。它能夠以百萬像素的分辨率渲染照片，得到逼真的視覺效果。迪士尼無疑是最適合這項工作的團隊。它們的目標是在不影響演員表演的同時，將該演員替換爲另一名演員的外貌。這非常具有挑戰性，但在很多情況下都很需要，比如改變角色的年齡，比如該演員不在的時候，比如對於演員來說太過危險的特技場景。目前的做法需要專業人員後期處理大量的逐幀動畫。"}]},{"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":"論文鏈接："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#7","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#7"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"8、更換自動編碼器的深度圖像處理 [8]"}]},{"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":"這種新技術可以改變任何圖片的紋理，同時使用完全無監督的訓練保持逼真效果！結果看起來甚至比 GANs 能實現的還要好，而且速度更快！它甚至可以用來製作真假難辨的贗品！"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/00\/007ba474ec3916f8cf05aaf991b7a281.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"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":"論文鏈接："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#8","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#8"}]}]},{"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":"代碼地址："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/rosinality\/swapping-autoencoder-pytorch?utm_source=catalyzex.com","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/rosinality\/swapping-autoencoder-pytorch?utm_source=catalyzex.com"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"9、GPT-3: 實現小樣本學習的語言模型 [9]"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/d4\/d41bc79d9e780999f11e2dcbf62f6b56.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"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":"目前最先進的自然語言處理系統難以泛化推廣以處理不同的任務。它們需要對數千個示例的數據集進行微調，而人類只需要看到幾個示例就可以執行新的語言任務。這就是 GPT-3 想要實現的目標，改進語言模型的任務無關特性。"}]},{"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":"論文鏈接："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#9","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#9"}]}]},{"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":"代碼地址："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/openai\/gpt-3","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/openai\/gpt-3"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"10、用於視頻修復的聯合時空變換 [10]"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/f7\/f71aed7bb81df8f71be11a25f0e0383f.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"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":"這種人工智能技術，可以填補刪除了移動物體之後缺失的像素，並且可以重建整個視頻。這種方法，比之前的方法都要更準確，更清晰。"}]},{"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":"論文鏈接："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#10","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#10"}]}]},{"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":"代碼地址："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/researchmm\/STTN?utm_source=catalyzex.com","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/researchmm\/STTN?utm_source=catalyzex.com"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"11、圖像 GPT——像素級別的生成預訓練 [11]"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/62\/62a8b0f3cbffa36bc05c35da839f74c8.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"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":"一個好的人工智能，比如 Gmail 中使用的那個，可以生成連貫的文本並補全你的短語。這張圖片使用了同樣的原則來完成一張圖片！所有這些都是在無人監督的訓練中完成的，根本不需要任何標籤！"}]},{"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":"論文鏈接："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#11","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#11"}]}]},{"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":"代碼地址："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/openai\/image-gpt","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/openai\/image-gpt"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"12、使用白盒卡通表示來學習卡通化 [12]"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/cf\/cfffadf7ac9fbdc860ae7ac5fc6f3da7.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"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":"輸入你想要的任意卡通風格，這個人工智能可以卡通化任何圖片或視頻！論文中列舉了很多示例，我們可以看看它是怎麼做到的。你甚至也可以像我一樣自己動手嘗試一下！"}]},{"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":"論文鏈接："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#12","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#12"}]}]},{"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":"代碼地址："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/SystemErrorWang\/White-box-Cartoonization","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/SystemErrorWang\/White-box-Cartoonization"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"13、FreezeG 凍結甄別器：用於微調 GAN 的一個簡單基準 [13]"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/21\/2108295a8473106d8b4f2a9cde85d7ef.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"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":"這個面部生成模型能夠將正常的面部照片轉換成獨特的風格，如 Lee malnyeon 的卡通風格，辛普森一家的風格，藝術風格，甚至是狗的風格！這項新技術最好的地方是它超級簡單，並且顯著優於以前在 GAN 中使用的技術。"}]},{"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":"論文鏈接："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#13","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#13"}]}]},{"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":"代碼地址："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/sangwoomo\/freezeD?utm_source=catalyzex.com","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/sangwoomo\/freezeD?utm_source=catalyzex.com"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"14、基於單個圖像中對人物重新進行神經式渲染 [14]"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/f3\/f39241c97ccc5370791dd38368c3db30.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"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":"該算法將人體姿態和形狀表示爲一個參數網格，可以從單一圖像中重構，且易於恢復。給定一個人的圖像，他們能夠根據另一個輸入圖像創建出這個人的不同姿勢或不同服裝的合成圖像。"}]},{"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":"論文鏈接："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#14","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#14"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"15、I2L-MeshNet：lixel 級預測網絡，用於從單一 RGB 圖像精確的 3D 人體姿態和網格預測 [15]"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/2a\/2a8bfc17d7a3954df4c459ca035260f4.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"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":"他們的目標是提出一種新的技術，用於從單一的 RGB 圖像 3D 人體姿態和網格預測。他們稱之爲 I2L-MeshNet。其中 I2L 表示圖像到 lixe 級。正如體素（體積 + 像素）是三維空間中的量化單元一樣，他們將 lixel（線和像素）定義爲一維空間中的量化單元。他們的方法比以前的方法性能更好，且代碼是公開的！"}]},{"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":"論文鏈接："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#15","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#15"}]}]},{"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":"代碼地址："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/mks0601\/I2L-MeshNet_RELEASE","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/mks0601\/I2L-MeshNet_RELEASE"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"16、超級導航圖：連續環境中的視覺語言導航 [16]"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/7b\/7b4584058a92386b47ff46bf4095a63c.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"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":"語言導航是一個被廣泛研究且非常複雜的領域。的確，對一個人來說，穿過一個房間拿你放在牀左邊牀頭櫃上的咖啡似乎很簡單。但對於智能體來說，情況就完全不同了。智能體是一種自主的人工智能驅動系統，該系統使用深度學習來執行任務。"}]},{"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":"論文鏈接："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#16","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#16"}]}]},{"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":"代碼地址："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/jacobkrantz\/VLN-CE","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/jacobkrantz\/VLN-CE"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"17、RAFT: 用於光流的遞歸全對場變換 [17]"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/5e\/5e62f9417a50afd14699e986baed6e0e.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"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":"此篇論文來自於普林斯頓大學的團隊，並獲得 ECCV 2020 最佳論文獎。研究者開發了一種新的端到端可訓練的光流模型。他們的方法超越了最先進的架構在多個數據集上的準確性，而且效率更高。"}]},{"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":"論文鏈接："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#17","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#17"}]}]},{"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":"代碼地址："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/princeton-vl\/RAFT","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/princeton-vl\/RAFT"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"18、衆包抽樣的全光功能 [18]"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/ee\/ee014965d2af21f22285bff04766f532.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"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":"利用遊客在網上公開的照片，他們能夠重建一個場景的多個視點，並保留真實的陰影和光線。對於全光場景渲染來說，這是一個巨大的進步，是當前最先進的技術。他們取得了令人驚歎的成績。"}]},{"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":"論文鏈接："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#18","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#18"}]}]},{"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":"代碼地址："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/zhengqili\/Crowdsampling-the-Plenoptic-Function","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/zhengqili\/Crowdsampling-the-Plenoptic-Function"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"19、通過深度潛在空間翻譯來恢復老照片 [19]"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/af\/af60cd8726728ef0b7dccb3e828cf34b.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"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":"想象一下，你祖母 18 歲時的老照片，充滿折皺，甚至可能被撕破過，如何能夠不露任何人工干預痕跡的高清照片呢？這就是所謂的舊照片修復，而這篇論文剛剛開闢了一個全新的途徑來解決這個問題，那就是使用深度學習的方法。"}]},{"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":"論文鏈接："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#19","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#19"}]}]},{"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":"代碼地址："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/microsoft\/Bringing-Old-Photos-Back-to-Life?utm_source=catalyzex.com","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/microsoft\/Bringing-Old-Photos-Back-to-Life?utm_source=catalyzex.com"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"20、支持可審計自治的神經迴路策略 [20]"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/5c\/5cb50d413129fe33e019876008886460.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"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":"來自奧地利 IST 和麻省理工學院的研究人員已經成功地使用一種新的人工智能系統訓練了一輛自動駕駛汽車，該系統基於小動物的大腦，比如蟯蟲。他們只用幾個神經元就能控制自動駕駛汽車，而流行的深度神經網絡如“啓蒙”(inveptions)、“重新網”(Resnets) 或 VGG 則需要數百萬個神經元。他們的網絡完全能夠控制一輛汽車，只需要使用由 19 個控制神經元組成的 75000 個參數，而不是數百萬個！"}]},{"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":"論文鏈接："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#20","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#20"}]}]},{"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":"代碼地址："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/mlech26l\/keras-ncp","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/mlech26l\/keras-ncp"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"21、返老？還童？[21]"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/33\/339b21713c39086515f38c49a1a279ec.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"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":"Adobe 研究院的一組研究人員開發了一種新的變換年齡的合成技術，該技術只基於一個人的一張照片，就它可以生成他任意年齡段的照片。"}]},{"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":"論文鏈接："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#21","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#21"}]}]},{"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":"代碼地址："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/royorel\/Lifespan_Age_Transformation_Synthesis","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/royorel\/Lifespan_Age_Transformation_Synthesis"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"22、DeOldify，黑白圖像着色技術 [22]"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/16\/16515f2d1139480d12f236800cabb5a1.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"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":"DeOldify 是一種對舊的黑白圖像甚至電影膠片進行着色和恢復的技術。它由 Jason Antic 開發，目前仍在更新中。這是現在給黑白圖像着色的最先進的方法，目前所有內容都是開源的。"}]},{"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":"論文鏈接："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#22","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#22"}]}]},{"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":"代碼地址："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/jantic\/DeOldify","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/jantic\/DeOldify"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"23、COOT: 用於視頻 - 文本表示學習的協作層次轉換器 [23]"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/20\/2089989273759f42fabf6031b0807a88.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"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":"顧名思義，它將視頻和對視頻的一般描述作爲輸入，使用 transformer 爲視頻的每一段生成準確的文本描述。"}]},{"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":"論文鏈接："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#23","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#23"}]}]},{"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":"代碼地址："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/gingsi\/coot-videotext","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/gingsi\/coot-videotext"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"24、有風格的神經畫師 [24]"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/cb\/cba6e8598e0a2cc4e24db384b4f63307.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"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":"這種圖像到繪畫的翻譯方法模擬了各種風格的真實畫家，它不像所有目前最先進的方法，而使用了一種不涉及任何 GAN 架構的新穎方法!"}]},{"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":"論文鏈接："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#24","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#24"}]}]},{"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":"代碼地址："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/jiupinjia\/stylized-neural-painting","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/jiupinjia\/stylized-neural-painting"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"25、實時人像摳圖真的需要綠幕嗎？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/49\/49f151faeaf599415c6981bbaf5d0bad.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"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":"人像摳圖是一項非常有趣的任務，目標是在照片中找到任何一個人，然後把背景去掉。由於任務的複雜性，這真的很難實現，這不得不雲找一個或多個擁有完美輪廓的人。在這篇論文中，作者回顧了多年來使用的最佳技術，以及 2020 年 11 月 29 日發表的一篇新方法。許多技術使用基礎的計算機視覺算法來實現這一任務，例如 GrabCut 算法，它非常快，但不是非常精確。"}]},{"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":"論文鏈接："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#25","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#25"}]}]},{"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":"代碼地址："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/ZHKKKe\/MODNet","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/ZHKKKe\/MODNet"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"26、ADA: 使用有限數據訓練生成對抗網絡 [26]"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/3a\/3a6b86259e10a27699773907c8ae57a5.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"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":"使用這種由 NVIDIA 開發的新的訓練方法，你可以用十分之一的圖像訓練一個強大的生成模型！"}]},{"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":"造福於許多拿不到這麼多圖像的應用程序！"}]},{"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":"論文鏈接："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#26","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#26"}]}]},{"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":"代碼地址："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/NVlabs\/stylegan2-ada","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/NVlabs\/stylegan2-ada"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"27、在立方球體上使用深度卷積神經網絡改進數據驅動的全球天氣預測 [27]"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/a0\/a0c78464ad8973001b53ad2eaf619f94.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"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":"目前傳統的天氣預報方法使用的是我們所謂的“數值天氣預報”模型。它使用大氣和海洋的數學模型來根據當前的情況預測天氣。它於 20 世紀 20 年代被首次引入，並在 20 世紀 50 年代通過計算機模擬產生仿真的結果。這些數學模型適用於進行短期和長期預測。但它的計算量很大，而且不能像深度神經網絡那樣基於那麼多數據進行預測。這正是這門技術如此有前途的部分原因。目前這些數值天氣預報模型已經使用機器學習作爲後處理工具來改進預報。天氣預報正受到機器學習研究人員越來越多的關注，已經產生了不錯的結果。"}]},{"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":"論文鏈接："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#27","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#27"}]}]},{"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":"代碼地址："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/jweyn\/DLWP-CS","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/jweyn\/DLWP-CS"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"28、NeRV：神經反射和可見域，用於重新照明和視圖合成 [28]"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/72\/72016f37bf961c2dab80eb1bacbc0f8c.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/29\/292e8345f342b2a96c74fb928c918f14.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"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":"這種新方法能夠生成一個完整的三維場景，並能夠決定場景中的照明。與以前的方法相比，所有這些都只需要非常有限的計算成本，但效果卻十分驚人。"}]},{"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":"論文鏈接："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#28","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/louisfb01\/Best_AI_paper_2020#28"}]}]},{"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":"代碼地址："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/people.eecs.berkeley.edu\/~pratul\/nerv\/","title":"","type":null},"content":[{"type":"text","text":"https:\/\/people.eecs.berkeley.edu\/~pratul\/nerv\/"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"總結"}]},{"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":"如你所見，對於人工智能來說，這是極具洞察力的一年，我超級興奮地想要看看 2021 年將會發生什麼！我們將持續跟蹤報道更多令人興奮的有趣論文，如果你對人工智能研究感興趣，歡迎保持關注！"}]},{"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":"引用論文"}]},{"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":"[1] A. Bochkovskiy, C.-Y. Wang, and H.-Y. M. Liao, Yolov4: Optimal speed and accuracy of object detection, 2020. arXiv:2004.10934 [cs.CV]."}]},{"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":"[2] S.-Y. Chen, W. Su, L. Gao, S. Xia, and H. Fu, “DeepFaceDrawing: Deep generation of face images from sketches,” ACM Transactions on Graphics (Proceedings of ACM SIGGRAPH2020), vol. 39, no. 4, 72:1–72:16, 2020."}]},{"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":"[3] S. W. Kim, Y. Zhou, J. Philion, A. Torralba, and S. Fidler, “Learning to Simulate DynamicEnvironments with GameGAN,” in IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Jun. 2020."}]},{"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":"[4] S. Menon, A. Damian, S. Hu, N. Ravi, and C. Rudin, Pulse: Self-supervised photo upsampling via latent space exploration of generative models, 2020. arXiv:2003.03808 [cs.CV]."}]},{"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":"[5] M.-A. Lachaux, B. Roziere, L. Chanussot, and G. Lample, Unsupervised translation of programming languages, 2020. arXiv:2006.03511 [cs.CL]."}]},{"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":"[6] S. Saito, T. Simon, J. Saragih, and H. Joo, Pifuhd: Multi-level pixel-aligned implicit function for high-resolution 3d human digitization, 2020. arXiv:2004.00452 [cs.CV]."}]},{"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] J. Naruniec, L. Helminger, C. Schroers, and R. Weber, “High-resolution neural face-swapping for visual effects,” Computer Graphics Forum, vol. 39, pp. 173–184, Jul. 2020.doi:10.1111\/cgf.14062."}]},{"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":"[8] T. Park, J.-Y. Zhu, O. Wang, J. Lu, E. Shechtman, A. A. Efros, and R. Zhang,Swappingautoencoder for deep image manipulation, 2020. arXiv:2007.00653 [cs.CV]."}]},{"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":"[9] T. B. Brown, B. Mann, N. Ryder, M. Subbiah, J. Kaplan, P. Dhariwal, A. Neelakantan, P.Shyam, G. Sastry, A. Askell, S. Agarwal, A. Herbert-Voss, G. Krueger, T. Henighan, R. Child, A. Ramesh, D. M. Ziegler, J. Wu, C. Winter, C. Hesse, M. Chen, E. Sigler, M. Litwin, S.Gray, B. Chess, J. Clark, C. Berner, S. McCandlish, A. Radford, I. Sutskever, and D. Amodei,“Language models are few-shot learners,” 2020. arXiv:2005.14165 [cs.CL]."}]},{"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":"[10] Y. Zeng, J. Fu, and H. Chao, Learning joint spatial-temporal transformations for video in-painting, 2020. arXiv:2007.10247 [cs.CV]."}]},{"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":"[11] M. Chen, A. Radford, R. Child, J. Wu, H. Jun, D. Luan, and I. Sutskever, “Generative pretraining from pixels,” in Proceedings of the 37th International Conference on Machine Learning, H. D. III and A. Singh, Eds., ser. Proceedings of Machine Learning Research, vol. 119, Virtual: PMLR, 13–18 Jul 2020, pp. 1691–1703. [Online]. Available:"},{"type":"link","attrs":{"href":"http:\/\/proceedings.mlr.press\/v119\/chen20s.html","title":"","type":null},"content":[{"type":"text","text":"http:\/\/proceedings.mlr.press\/v119\/chen20s.html"}]},{"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":"[12] Xinrui Wang and Jinze Yu, “Learning to Cartoonize Using White-box Cartoon Representations.”,IEEE Conference on Computer Vision and Pattern Recognition, June 2020."}]},{"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":"[13] S. Mo, M. Cho, and J. Shin, Freeze the discriminator: A simple baseline for fine-tuning gans,2020. arXiv:2002.10964 [cs.CV]."}]},{"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":"[14] K. Sarkar, D. Mehta, W. Xu, V. Golyanik, and C. Theobalt, “Neural re-rendering of humans from a single image,” in European Conference on Computer Vision (ECCV), 2020."}]},{"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":"[15] G. Moon and K. M. Lee, “I2l-meshnet: Image-to-lixel prediction network for accurate 3d human pose and mesh estimation from a single rgb image,” in European Conference on ComputerVision (ECCV), 2020"}]},{"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":"[16] J. Krantz, E. Wijmans, A. Majumdar, D. Batra, and S. Lee, “Beyond the nav-graph: Vision-and-language navigation in continuous environments,” 2020. arXiv:2004.02857 [cs.CV]."}]},{"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":"[17] Z. Teed and J. Deng, Raft: Recurrent all-pairs field transforms for optical flow, 2020. arXiv:2003.12039 [cs.CV]."}]},{"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":"[18] Z. Li, W. Xian, A. Davis, and N. Snavely, “Crowdsampling the plenoptic function,” inProc.European Conference on Computer Vision (ECCV), 2020."}]},{"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":"[19] Z. Wan, B. Zhang, D. Chen, P. Zhang, D. Chen, J. Liao, and F. Wen, Old photo restoration via deep latent space translation, 2020. arXiv:2009.07047 [cs.CV]."}]},{"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":"[20] Lechner, M., Hasani, R., Amini, A.et al.Neural circuit policies enabling auditable autonomy.Nat Mach Intell2,642–652 (2020)."},{"type":"link","attrs":{"href":"https:\/\/doi.org\/10.1038\/s42256-020-00237-3","title":"","type":null},"content":[{"type":"text","text":"https:\/\/doi.org\/10.1038\/s42256-020-00237-3"}]}]},{"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":"[21] R. Or-El, S. Sengupta, O. Fried, E. Shechtman, and I. Kemelmacher-Shlizerman, “Lifespanage transformation synthesis,” in Proceedings of the European Conference on Computer Vision(ECCV), 2020."}]},{"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":"[22] Jason Antic, Creator of DeOldify,"},{"type":"link","attrs":{"href":"https:\/\/github.com\/jantic\/DeOldify","title":"","type":null},"content":[{"type":"text","text":"https:\/\/github.com\/jantic\/DeOldify"}]}]},{"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":"[23] S. Ging, M. Zolfaghari, H. Pirsiavash, and T. Brox, “Coot: Cooperative hierarchical trans-former for video-text representation learning,” in Conference on Neural Information ProcessingSystems, 2020."}]},{"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":"[24] Z. Zou, T. Shi, S. Qiu, Y. Yuan, and Z. Shi, Stylized neural painting, 2020. arXiv:2011.08114[cs.CV]."}]},{"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":"[25] Z. Ke, K. Li, Y. Zhou, Q. Wu, X. Mao, Q. Yan, and R. W. Lau, “Is a green screen really necessary for real-time portrait matting?” ArXiv, vol. abs\/2011.11961, 2020."}]},{"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":"[26] T. Karras, M. Aittala, J. Hellsten, S. Laine, J. Lehtinen, and T. Aila, Training generative adversarial networks with limited data, 2020. arXiv:2006.06676 [cs.CV]."}]},{"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":"[27]J. A. Weyn, D. R. Durran, and R. Caruana, “Improving data-driven global weather prediction using deep convolutional neural networks on a cubed sphere”, Journal of Advances in Modeling Earth Systems, vol. 12, no. 9, Sep. 2020, issn: 1942–2466.doi:10.1029\/2020ms002109"}]},{"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":"[28] P. P. Srinivasan, B. Deng, X. Zhang, M. Tancik, B. Mildenhall, and J. T. Barron, “Nerv: Neural reflectance and visibility fields for relighting and view synthesis,” in arXiv, 2020."}]},{"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":"原文鏈接："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https:\/\/medium.com\/towards-artificial-intelligence\/2020-a-year-full-of-amazing-ai-papers-a-review-c42fa07aff4b","title":"","type":null},"content":[{"type":"text","text":"https:\/\/medium.com\/towards-artificial-intelligence\/2020-a-year-full-of-amazing-ai-papers-a-review-c42fa07aff4b"}]}]},{"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":"譯者簡介：冬雨，小小技術宅一枚，從事研發過程改進及質量改進方面的工作，關注編程、軟件工程、敏捷、DevOps、雲計算等領域，非常樂意將國外新鮮的 IT 資訊和深度技術文章翻譯分享給大家，已翻譯出版《深入敏捷測試》、《持續交付實戰》。"}]}]}