Chung-Yi Weng 研究

PersonNeRF: Personalized Reconstruction from Photo Collections PersonNeRF:  照片集的個性化重建   (CVPR 2023 accepted papers ) https://grail.cs.washington.edu/projects/personnerf/   arXiv:2302.08504 (cs) [Submitted on 16 Feb 2023] Chung-Yi Weng, Pratul P. Srinivasan, Brian Curless, Ira Kemelmacher-Shlizerman   We present PersonNeRF, a method that takes a collection of photos of a subject (e.g. Roger Federer) captured across multiple years with arbitrary body poses and appearances, and enables rendering the subject with arbitrary novel combinations of viewpoint, body pose, and appearance. PersonNeRF builds a customized neural volumetric 3D model of the subject that is able to render an entire space spanned by camera viewpoint, body pose, and appearance. A central challenge in this task is dealing with sparse observations; a given body pose is likely only observed by a single viewpoint with a single appearance, and a given appearance is only observed under a handful of different body poses. We add

  NeRF 擺脫了“動畫照片”的束縛  (  NeRF Breaks Free From Being an ‘Animated Photo’) 本篇介紹  PersonNeRF: Personalized Reconstruction from Photo Collections ( CVPR 2023 accepted papers ) 英國 AI/VR 初創公司 Metaphysic   報導 https://metaphysic.ai/   https://metaphysic.ai/nerf-breaks-free-from-being-an-animated-photo/     NeRF Breaks Free From Being an NeRF 擺脫了“動畫照 片”的束縛 February 17, 2023 1:26 pm ABOUT THE AUTHOR   馬丁·安德森(  Martin Anderson)  報導  : 我是 Martin Anderson，一位專注於機器學習、人工智能、大數據和密切相關主題的作家，重點是圖像合成、計算機視覺和 NLP。 隨著當前圍繞穩定擴散的所有喧囂，神經輻射場 (NeRF) 最近並沒有得到太多的愛。雖然它算不上是一種“遺留”技術（它在 2020 年才出現），但與潛在擴散網絡甚至生成對抗網絡的豐富且可探索的潛在空間相比，它可以說是一種神經渲染的笨拙方法。甘）。 NeRF 難以編輯，難以包含在 deepfakes 管道中，並且比人類合成更適合城市場景和靜態表示，儘管如此，NeRF 可能仍然是目前可用於人類形式的最準確的神經表示技術——但它並不是最有想像力的。 與 Stable Diffusion 不同，您無法“搜索”NeRF 的潛在空間以尋找隱藏的寶藏，因為 NeRF 表示幾乎僅限於它從中導出其網絡的照片中存在的任何材料。如果該材料是工作室中的女性，您將無法干預從這些源圖像生成的訓練模型（至少不能以任何簡單或有意義的方式進行干預）。用侏羅紀公園 (1993) 中的約翰·哈蒙德 (John Hammond) 的話說，NeRF 是“一種騎行”。 NVIDIA 的 InstantNeRF 僅從四張圖像推斷出令人印象深刻的面部視圖範圍，但分辨率、表情準確性和移動性仍然是高分辨率推理的主要挑戰。  資料來源：https

    PersonNeRF : Personalized Reconstruction from Photo Collections ( CVPR 2023 ) Chung-Yi Weng  1 Pratul P. Srinivasan  2 Brian Curless  1,2 Ira Kemelmacher-Shlizerman  1,2   1 University of Washington 2 Google Research         PersonNeRF builds a personalized space from photos of Roger Federer, rendering with novel combinations of viewpoint, body pose, and appearance.  Try a laptop browser if the video can't load Abstract We present PersonNeRF, a method that takes a collection of photos of a subject (e.g., Roger Federer) captured across multiple years with arbitrary body poses and appearances, and enables rendering the subject with arbitrary novel combinations of viewpoint, body pose, and appearance. PersonNeRF builds a customized neural volumetric 3D model of the subject that is able to render an entire space spanned by camera viewpoint, body pose, and appearance. A central challenge in this task is dealing with sparse observations; a given body