BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:America/Los_Angeles BEGIN:DAYLIGHT DTSTART:20230312T030000 TZOFFSETFROM:-0800 TZOFFSETTO:-0700 RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3 TZNAME:PDT END:DAYLIGHT BEGIN:STANDARD DTSTART:20221106T010000 TZOFFSETFROM:-0700 TZOFFSETTO:-0800 RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11 TZNAME:PST END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20221208T211935Z UID:23C171C0-34FC-4C76-8B5A-29C0B877A666 DTSTART;TZID=America/Los_Angeles:20221207T183000 DTEND;TZID=America/Los_Angeles:20221207T200000 DESCRIPTION:Autonomous robots depend on their perception systems to underst and the world around them. These machines often leverage a host of sensors including cameras\, lidars\, radars\, and ultrasonic sensors to create th is environmental understanding. Stereo cameras play a big role in providin g depth perception to robotic systems. This depth information can be estim ated using classical computer vision techniques\, like semi-global matchin g (SGM) or leverage deep neural networks (DNNs). Each individual algorithm may struggle in a particular set of operating conditions. But when multip le depth estimation algorithms are leveraged simultaneously\, It is possib le that more robust depth information can be calculated.\n\nIn this talk\, we'll cover work at NVIDIA to train the ESS DNN model for determining ste reo disparity using both synthetic and real-world data to perform well whe re SGM may not. We'll also introduce the Bi3D model which is trained on th e simplified question of "is X closer than M meters?" rather than "how far away is X?"\, yielding improvements in both accuracy and speed. As every approach has deficiencies on its own\, we'll touch upon how ensembling the responses of ESS and Bi3D\, DNNs developed specifically for robotic perce ption with SGM could lead to robust obstacle detection. Finally\, we'll di scuss how we've tuned the performance of these models to run on embedded c ompute for the responsive stopping behavior required in autonomous mobile robots (AMRs).\n\nSpeaker(s): Hemal Shah\, Gerard Andrews\n\nAgenda: \n6:3 0 PM Introduction (Tom Coughlin)\n\n6:45 PM Talk\n\n7:30 PM Q&A\n\n8:00 PM End\n\nVirtual: https://events.vtools.ieee.org/m/332953 LOCATION:Virtual: https://events.vtools.ieee.org/m/332953 ORGANIZER:tom@tomcoughlin.com SEQUENCE:14 SUMMARY:Robust depth estimation for robots with stereo cameras using bespok e DNNs URL;VALUE=URI:https://events.vtools.ieee.org/m/332953 X-ALT-DESC:Description: <br /><p style="font-weight: 400\;">Autonomous robo ts depend on their perception systems to understand the world around them. These machines often leverage a host of sensors including cameras\, lidar s\, radars\, and ultrasonic sensors to create this environmental understan ding. Stereo cameras play a big role in providing depth perception to robo tic systems. This depth information can be estimated using classical compu ter vision techniques\, like semi-global matching (SGM) or leverage deep n eural networks (DNNs). Each individual algorithm may struggle in a particu lar set of operating conditions. But when multiple depth estimation algori thms are leveraged simultaneously\, It is possible that more robust depth information can be calculated.</p>\n<p style="font-weight: 400\;">In this talk\, we'll cover work at NVIDIA to train the ESS DNN model for determini ng stereo disparity using both synthetic and real-world data to perform we ll where SGM may not. We'll also introduce the Bi3D model which is trained on the simplified question of "is X closer than M meters?" rather than "h ow far away is X?"\, yielding improvements in both accuracy and speed. As every approach has deficiencies on its own\, we'll touch upon how ensembli ng the responses of ESS and Bi3D\, DNNs developed specifically for robotic perception with SGM could lead to robust obstacle detection. Finally\, we 'll discuss how we've tuned the performance of these models to run on embe dded compute for the responsive stopping behavior required in autonomous m obile robots (AMRs).</p><br /><br />Agenda: <br /><p>6:30 PM Introduction (Tom Coughlin)</p>\n<p>6:45 PM Talk</p>\n<p>7:30 PM Q&amp\;A</p>\n<p>8:00 PM End</p> END:VEVENT END:VCALENDAR