BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Beyond Visual Geometry: Toward Physical 3D Reconstruction - Chuanx
 ia Zheng\, University of Oxford and Nanyang Technological University
DTSTART:20250724T100000Z
DTEND:20250724T110000Z
UID:TALK234463@talks.cam.ac.uk
CONTACT:Elliott Wu
DESCRIPTION:Recent progress in 3D Visual Geometry has led to impressive re
 construction and generation results using purely feed-forward deep network
 s. However\, much of this progress remains limited to geometric surface mo
 deling or photorealistic novel view synthesis\, often overlooking crucial 
 physical attributes such as occluded geometry\, physical stability\, and d
 ynamic motion. In this talk\, I will present a series of work that go beyo
 nd visible geometry. I will introduce Amodal3R\, which defines amodal 3D r
 econstruction from occluded 2D images\, predicting complete 3D geometry an
 d appearance when significant portions are hidden from views. Next\, I wil
 l discuss DSO\, which aligns 3D generative models with physics based simul
 ation feedback\, ensuring reconstruction are not just visually plausible b
 ut physically sound under gravity. To this end\, I will introduce Geo4D\, 
 to explore how one can build high-quality 4D reconstruction networks start
 ing form video generators pre-trained on millions of videos.\n\n[1] Wu\, T
 .\, Zheng\, C.\, Guan\, F.\, Vedaldi\, A.\, & Cham\, T. J\, “Amodal3r: A
 modal 3d reconstruction from occluded 2d images”. ICCV 2025.\n[2] Li\, R
 .\, Zheng\, C.\, Rupprecht\, C.\, & Vedaldi\, A. “DSO: Aligning 3D Gener
 ators with Simulation Feedback for Physical Soundness”. ICCV 2025.\n[3] 
 Jiang\, Z.\, Zheng\, C.\, Laina\, I.\, Larlus\, D.\, & Vedaldi\, A. Geo4d:
  Leveraging video generators for geometric 4d scene reconstruction. ICCV 2
 025.\n\nSpeaker Bio:\nChuanxia Zheng is an incoming Nanyang Assistant Prof
 essor at Nanyang Technological University (NTU)\, where he will lead the p
 hysical vision group. He is currently a MSCA Postdoctoral fellow in Visual
  Geometry Group (VGG) at the University of Oxford\, working with Andrea Ve
 daldi on feed-forward photorealistic 3D and 4D reconstruction. He received
  his Ph.D. degree from NTU.\n\nHis research focuses on Creative AI\, with 
 the goal of developing systems that can perceive\, reconstruct and interac
 t with the physical world. The broader goal is to create realistic digital
  twins of the physical natural world\, capturing diverse physical properti
 es\, such as geometry\, appearance\, occlusion\, motion\, interaction\, gr
 avity\, sound and more.
LOCATION: Cambridge University Engineering Department\, JDB Teaching Room
END:VEVENT
END:VCALENDAR