← 研踪 PaperTrailCVF · CVPR 2026:cvf:f845c9bd522aaa0c2af4b9f8bd30e936 · 记录 #1
SAM 3D: 3Dfy Anything in Images
Chen, XingyuCHU, FU-JENGleize, PierreLiang, Kevin JSax, AlexanderTang, HaoWang, WeiyaoGuo, MichelleHardin, ThibautLi, XiangLin, AohanLiu, Jia-WeiMa, ZiqiSagar, AnushkaSong, BowenWang, XiaodongYang, JianingZhang, BowenDollár, PiotrGkioxari, GeorgiaFeiszli, MattMalik, Jitendra 来源日期（仅到年份） 2026
来源记录日期 2026
雷达观测 2026-11-02 00:05
摘要
We present SAM 3D, a generative model for visually grounded 3D object reconstruction, predicting geometry, texture, and layout from a single image. SAM 3D excels in natural images, where occlusion and scene clutter are common and visual recognition cues from context play a larger role. We achieve this with a human- and model-in-the-loop pipeline for annotating object shape, texture, and pose, providing visually grounded 3D reconstruction data at unprecedented scale. We learn from this data in a modern, multi-stage training framework that combines synthetic pretraining with real-world alignment, breaking the 3D "data barrier". We obtain significant gains over recent work, with at least a 5:1 win rate in human preference tests on real-world objects and scenes. We will release our code and model weights, an online demo, and a new challenging benchmark for in-the-wild 3D object reconstruction.
论文发现出处
这些页面提供论文线索，不代表同行评审或质量认证。
观测到的版本
平台记录号代表本站观测到的元数据变化，不是作者修订号。来源日期仅到月份或年份时，按该月或该年首日存档。
记录 #1 · 2026 · 最新观测版本
SAM 3D: 3Dfy Anything in Images
We present SAM 3D, a generative model for visually grounded 3D object reconstruction, predicting geometry, texture, and layout from a single image. SAM 3D excels in natural images, where occlusion and scene clutter are common and visual recognition cues from context play a larger role. We achieve this with a human- and model-in-the-loop pipeline for annotating object shape, texture, and pose, providing visually grounded 3D reconstruction data at unprecedented scale. We learn from this data in a modern, multi-stage training framework that combines synthetic pretraining with real-world alignment, breaking the 3D "data barrier". We obtain significant gains over recent work, with at least a 5:1 win rate in human preference tests on real-world objects and scenes. We will release our code and model weights, an online demo, and a new challenging benchmark for in-the-wild 3D object reconstruction.
查看此版本解读 →查看此版本原文 ↗