BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Asia/Tokyo
X-LIC-LOCATION:Asia/Tokyo
BEGIN:STANDARD
TZOFFSETFROM:+0900
TZOFFSETTO:+0900
TZNAME:JST
DTSTART:18871231T000000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260817T171531Z
LOCATION:Hall B7 (1)\, B Block\, Level 7
DTSTART;TZID=Asia/Tokyo:20241204T150800
DTEND;TZID=Asia/Tokyo:20241204T151900
UID:siggraphasia_SIGGRAPH Asia 2024_sess120_papers_865@linklings.com
SUMMARY:A Plentoptic 3D Vision System
DESCRIPTION:Agastya Kalra, Vage Tamaazyan, Alberto Dall'olio, Raghav Khann
 a, Tomas Gerlich, Georgia Giannopolou, Guy Stoppi, Daniel Baxter, and Abhi
 jit Ghosh (Intrinsic); Rick Szeliski (Google Research); and Kartik Venkata
 raman (Intrinsic)\n\nWe present a novel multi-camera, multi-modal vision s
 ystem designed for industrial robotics applications. The system generates 
 high-quality 3D point clouds, with a focus on improving the completeness a
 nd reducing hallucinations for collision avoidance across various geometri
 es, materials, and lighting conditions. Our system incorporates several ke
 y advancements: (1) a modular and scalable \textbf{Plenoptic Stereo Vision
  Unit} that captures high-resolution RGB, polarization, and infrared (IR) 
 data for enhanced scene understanding; (2) an \textbf{Auto-Calibration Rou
 tine} that enables the seamless addition and automatic registration of mul
 tiple stereo units, expanding the system's capabilities; (3) a \textbf{Dee
 p Fusion Stereo Architecture} - a state-of-the-art deep learning architect
 ure trained fully on synthetic data that effectively fuses multi-baseline 
 and multi-modal data for superior reconstruction accuracy. We demonstrate 
 the impact of each design decision through rigorous testing, showing impro
 ved performance across varying lighting, geometry, and material challenges
 . To benchmark our system, we create an extensive industrial-robotics insp
 ired dataset featuring sub-millimeter accurate ground truth 3D reconstruct
 ions of scenes with challenging elements such as sunlight, deep bins, tran
 sparency, reflective surfaces, and thin objects. Our system surpasses the 
 performance of state-of-the-art high-resolution structured light on this d
 ataset. We also demonstrate generalization to non-robotics polarization da
 tasets. Interactive visualizations and videos are available at \url{https:
 //www.intrinsic.ai/publications/siggraphasia2024}.\n\nRegistration Categor
 y: Full Access, Full Access Supporter\n\nLanguage Format: English Language
 \n\nSession Chair: Hao (Richard) Zhang (Simon Fraser University, Augmenta)
 \n\n
URL:https://asia.siggraph.org/2024/program/?id=papers_865&sess=sess120
END:VEVENT
END:VCALENDAR