BEGIN:VCALENDAR

VERSION:2.0

PRODID:-//wp-events-plugin.com//6.6.4.4//EN

TZID:Asia/Jerusalem

X-WR-TIMEZONE:Asia/Jerusalem
BEGIN:VEVENT

UID:0-263@aerospace.technion.ac.il

DTSTART;TZID=Asia/Jerusalem:20210322T123000

DTEND;TZID=Asia/Jerusalem:20210322T133000

DTSTAMP:20230525T063708Z

URL:https://aerospace.technion.ac.il/events/autonomous-classification-unde
 r-uncertainty/

SUMMARY:Autonomous Classification Under Uncertainty
DESCRIPTION:Lecturer:Vladimir Tchuiev\n Faculty:Department of Aerospace Eng
 ineering\n Institute:Technion – Israel Institute of Technology\n Locatio
 n:https://technion.zoom.us/j/95387827267\n Zoom: \n Abstract: \n Details: 
 \n Classification and object recognition is a fundamental problem in many 
 robotics and aerospace applications\, such as autonomous driving\, vision-
 based navigation\, and search &amp\; rescue.\nThe field has advanced much 
 in recent years with the introduction of deep-learning-based approaches\, 
 yet reliable classification remains a significant problem. Classification 
 results may be\naffected by varying viewpoints\, changing lighting conditi
 ons\, occlusions\, localization uncertainty\, and limited by the classifie
 r's training set. In this work\, we propose several sequential classificat
 ion approaches that deal with some of these uncertainties within a semanti
 c simultaneous localization and mapping (SLAM) framework.\nFirst\, we prop
 ose using a viewpoint-dependent classifier model\, which uses the coupling
  between object class and pose to assist in addressing classification and 
 perceptual aliasing. We do so by maintaining a hybrid belief over continuo
 us and discrete random variables. One robot may prove insufficient in clas
 sifying the objects within the environment\, so we propose a formulation t
 hat uses the viewpoint-dependent model within a distributed multi-robot se
 tting\, while keeping the estimation consistent for both continuous and di
 screte random variables. Furthermore\,\nthe classifier's training set is l
 imited\, and during deployment the robot may encounter scenarios in which 
 it was not trained on\, inducing epistemic uncertainty. We propose a seque
 ntial classification approach that accounts for posterior epistemic uncert
 ainty from a sequence of images. Eventually\, we incorporate posterior epi
 stemic uncertainty within a belief space planning (BSP) framework\, consid
 ering in particular autonomous classification and active semantic SLAM. We
  study our approaches in simulation and using real data.\nZoom Meeting
CATEGORIES:Seminars
LOCATION:https://technion.zoom.us/j/95387827267

END:VEVENT

BEGIN:VTIMEZONE

TZID:Asia/Jerusalem

X-LIC-LOCATION:Asia/Jerusalem

BEGIN:STANDARD

DTSTART:20201025T010000

TZOFFSETFROM:+0300

TZOFFSETTO:+0200

TZNAME:IST

END:STANDARD

END:VTIMEZONE
END:VCALENDAR