BEGIN:VCALENDAR
VERSION:2.0
PRODID:IEEE vTools.Events//EN
CALSCALE:GREGORIAN
BEGIN:VTIMEZONE
TZID:Europe/Berlin
BEGIN:DAYLIGHT
DTSTART:20260329T030000
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3
TZNAME:CEST
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20251026T020000
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10
TZNAME:CET
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260329T195014Z
UID:14C6ED52-7E7C-48C2-95F1-A143E9A871F3
DTSTART;TZID=Europe/Berlin:20260326T100000
DTEND;TZID=Europe/Berlin:20260326T113000
DESCRIPTION:Vision models now match or exceed human accuracy on many benchm
 arks\, yet they can fail for the “wrong reasons”: brittle decision bou
 ndaries\, attention that diverges from human strategies\, and invariances 
 that do not reflect perception. In this talk\, I present a unified researc
 h program toward trustworthy\, human-aligned vision\, organized around a s
 imple question: what does a model treat as evidence\, and how stable is th
 at evidence under change?\n\nSpeaker(s): Dr. Dario Zanca\n\nRoom: 00.014 (
 Seminarraum 1)\, EG\, Department Artificial Intelligence in Biomedical Eng
 ineering (AIBE)\, FAU Erlangen-Nuremberg\, Nuernberger Str. 74\, Erlangen\
 , Bayern\, Germany\, 91052\, Virtual: https://events.vtools.ieee.org/m/543
 051
LOCATION:Room: 00.014 (Seminarraum 1)\, EG\, Department Artificial Intellig
 ence in Biomedical Engineering (AIBE)\, FAU Erlangen-Nuremberg\, Nuernberg
 er Str. 74\, Erlangen\, Bayern\, Germany\, 91052\, Virtual: https://events
 .vtools.ieee.org/m/543051
ORGANIZER:alexandros_tanzanakis@ieee.org
SEQUENCE:189
SUMMARY:Invited Talk: Toward Trustworthy\, Human-Aligned Vision (Dr. Dario 
 Zanca\, FAU Erlangen-Nuremberg)
URL;VALUE=URI:https://events.vtools.ieee.org/m/543051
X-ALT-DESC:Description: <br /><p><span style="font-size: 12.0pt\; font-fami
 ly: 'Calibri'\,sans-serif\; mso-fareast-font-family: 'Times New Roman'\; c
 olor: black\; mso-ansi-language: #1000\; mso-fareast-language: #1000\; mso
 -bidi-language: AR-SA\;">Vision models now match or exceed human accuracy 
 on many benchmarks\, yet they can fail for the &ldquo\;wrong reasons&rdquo
 \;: brittle decision boundaries\, attention that diverges from human strat
 egies\, and invariances that do not reflect perception. In this talk\, I p
 resent a unified research program toward trustworthy\, human-aligned visio
 n\, organized around a simple question: what does a model treat as evidenc
 e\, and how stable is that evidence under change?</span></p>
END:VEVENT
END:VCALENDAR