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:20250110T023312Z
LOCATION:Hall B5 (2)\, B Block\, Level 5
DTSTART;TZID=Asia/Tokyo:20241205T111300
DTEND;TZID=Asia/Tokyo:20241205T112700
UID:siggraphasia_SIGGRAPH Asia 2024_sess128_papers_618@linklings.com
SUMMARY:High-Throughput Batch Rendering for Embodied AI
DESCRIPTION:Technical Papers\n\nLuc Guy Rosenzweig, Brennan Shacklett, War
 ren Xia, and Kayvon Fatahalian (Stanford University)\n\nIn this paper we s
 tudy the problem of efficiently rendering images for embodied AI training 
 workloads, where agent training involves rendering millions to billions of
  independent frames, often at low-resolutions and with simple (or no) ligh
 ting and shading, that serve as the agent's observations of the world. To 
 enable high-throughput end-to-end training, we design, and provide a high-
 performance GPU implementation of, a frontend renderer interface that allo
 ws state-of-the-art GPU-accelerated batch world simulators to efficiently 
 communicate with high-performance rendering backends for generating agent 
 observations. Using this interface we architect and compare two high-perfo
 rmance renderers: one based on the GPU hardware-accelerated graphics pipel
 ine and a second based on a GPU software implementation of ray tracing.To 
 evaluate these renderers and encourage further research by the graphics co
 mmunity in this area, we build a rendering benchmark for this underexplore
 d regime and find that the ray tracing based renderer outperforms the rast
 erization based solution across the benchmark on a datacenter class GPU, w
 hile also performing competitively in geometrically complex environments o
 n a high-end consumer GPU. When tasked to render large batches of independ
 ent 128x128 images, the raytracer can exceed 100,000 frames per second per
  GPU for simple scenes, and exceed 10,000 frames per second per GPU on geo
 metrically complex scenes from the HSSD dataset.\n\nRegistration Category:
  Full Access, Full Access Supporter\n\nLanguage Format: English Language\n
 \nSession Chair: Manolis Savva (Simon Fraser University)
URL:https://asia.siggraph.org/2024/program/?id=papers_618&sess=sess128
END:VEVENT
END:VCALENDAR