BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:IC Colloquium: Compositional Perception for Action
DTSTART:20190404T101500
DTEND:20190404T111500
DTSTAMP:20260407T051055Z
UID:8bc9a72b4183401a1ba7c1bd433a36d608347e831803eceed4d9b697
CATEGORIES:Conferences - Seminars
DESCRIPTION:By: Amir Zamir - Stanford University and UC Berkeley\nIC Facul
 ty candidate\n\nAbstract:\nArtificial Intelligence seeks agents that can p
 erceive the world and act accordingly. Despite remarkable progress toward 
 this goal\, a fundamental shortcoming persists on the perception front: di
 fficulty in scaling to the complexity of the real world\, and consequently
 \, reducing the operation domain to perceptually simplified ones (e.g. con
 trolled spaces\, video games\, tabletop manipulation scenarios). I will pr
 esent my efforts toward a visual perception that can ultimately scale to r
 eal-world complexity and support goals of active agents by going beyond is
 olated pattern recognition problems. The core idea is utilizing compositio
 nality to tame the complexity: the world is largely composed of shared vis
 ual factors\, hence a compact perceptual skill set can be sufficient for u
 nderstanding large parts of it. I will show a method for tractably learnin
 g a large set of perception tasks using transfer learning (Taskonomy)\, to
 ward forming a multi-task compositional perception dictionary. I will show
  this dictionary can be turned into a mid-level perception module for acti
 ve robotic agents\, enabling them to perceive in the real world and improv
 e their sample efficiency and generalization. This is accomplished using b
 oth real robots as well as a virtual environment rooted in real spaces (Gi
 bson Environment). I will conclude with discussing cross-task consistency 
 and employing that as an intrinsic source of supervision for continual fin
 e tuning of a compositional perception.  \n\nBio:\nAmir Zamir is a postdo
 ctoral researcher at Stanford University and University of California\, Be
 rkeley. His research interests are broadly in computer vision and machine 
 learning with a focus on transfer/self/un supervised learning and percepti
 on-for-robotics. He has been recognized with CVPR (2018) Best Paper Award\
 , CVPR (2016) Best Student Paper Award\, NVIDIA Pioneering Research Award 
 (2018)\, and Stanford ICME Seed Award (2016)\, among others. His research 
 has been covered by popular press outlets\, such as NPR or The New York Ti
 mes.\n\nMore information
LOCATION:BC 420 https://plan.epfl.ch/?room==BC%20420
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR