道格•羅勃: 看似真人的數位人類

看似真人的數位人類

在這場精彩的演說和技術展示中，軟體研究者道格•羅勃讓「數位道格」初次登台：他是道格的即時 3D 數位呈現，精確程度到了連毛孔和皺紋都一模一樣。在慣性動作捕捉衣、深度神經網路、和大批資料的協助之下，數位道格能呈現出真實道格的情緒（甚至他的血流和睫毛開合）到極細緻的程度。來聽聽這場演說，進一步了解這項令人驚艷的技術是如何打造出來的，以及能如何將它應用在電影、虛擬助理和更多其他的領域中。

Advances in machine learning and motion capture now make it possible to create digital humans that mimic real people in astonishing detail, from skin pores to blood flow. A single camera and a motion capture suit can drive a 3D avatar live, capturing facial micro-expressions and subtle gestures in just 16 milliseconds. This technology emerged from one of the largest facial data sets ever recorded and promises to reshape film, virtual assistants, and online communication. While the results are striking, they also raise new questions about trust and authenticity in a world where appearances can be instantly changed.