【精英翻譯】的部落格

AI can render 3D objects from 2D pictures 人工智慧可用二維平面照片算出三維立體物體

精英翻譯社轉自http://iservice.ltn.com.tw/Service/english/english.php?engno=1211530&day=2018-06-25

◎ 劉宜庭

Google subsidiary DeepMind unveiled a new type of computer vision algorithm that can generate 3D models of a scene from 2D snapshots：the Generative Query Network （GQN）.

Google（谷歌）的子公司DeepMind揭露一款新的電腦視覺演算法：「生成式查詢網路（GQN）」，可從二維平面照片生成三維立體場景模型。

The GQN, details of which were published in Science, can “imagine” and render scenes from any angle without any human supervision or training.

「生成式查詢網路」的詳細資訊發表在《科學》期刊，它可以「想像」並從任何角度呈現場景，不需要任何人類的監督或訓練。

The two-part system is made up of a representation network and a generation network. The former takes input data and translates it into a mathematical representation（a vector）describing the scene, and the latter images the scene.

這套分為兩部分的系統，由表徵網路和生成網路組成。前者可將輸入的數據資料轉換成用來描述場景的數學陳述（向量資料），後者可將該場景圖像化。

To train the system, DeepMind researchers fed GQN images of scenes from different angles, which it used to teach itself about the textures, colors, and lighting of objects independently of one another and the spatial relationships between them. It then predicted what those objects would look like off to the side or from behind.

為了訓練這套系統，DeepMind提供場景中各種不同角度的影像給「生成式查詢網路」，讓它自學各個物體的材質、顏色和亮度，以及物體間的空間關係。該系統便可推算出這些物體看起來更像側邊或後面。

Using its spatial understanding, the GQN could control the objects. And it self-corrects as it moves around the scene, adjusting its predictions when they prove incorrect.

透過對空間的理解，「生成式查詢網路」可掌握各個物體。而且，它還能在場景中移動時進行自我修正，在證實錯誤時校正其推算式。