BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Visual Understanding without Naming:Bypassing the “Language Bott
 leneck”
DTSTART:20150619T100000
DTEND:20150619T110000
DTSTAMP:20260410T114245Z
UID:dd0e261fde3b44938a4954d2f4c7d3ff3953c9e001e285d73e039906
CATEGORIES:Conferences - Seminars
DESCRIPTION:Alexei (Alyosha) Efros\, UC Berkeley (USA)\nMost modern visual
  understanding approaches rely on supervision by\nword labels to achieve t
 heir impressive performance.  But there are\nmany more things in our visu
 al world than we have words to describe\nthem with.\nUsing language is sup
 ervisory signal risks missing out on much of the\nvisual subtlety.  In th
 is talk\, I will describe some of our recent\nefforts to bypass the "langu
 age bottleneck" and instead use\ninformation that is already in the data\,
  such as spatial context and\nvisual consistency\, to help in visual under
 standing\, visual\ncorrespondence\, and image retrieval.
LOCATION:BC 420 https://plan.epfl.ch/?room==BC%20420
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR