BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//wp-events-plugin.com//7.2.3.1//EN
TZID:Asia/Kolkata
X-WR-TIMEZONE:Asia/Kolkata
BEGIN:VEVENT
UID:136@cds.iisc.ac.in
DTSTART;TZID=Asia/Kolkata:20250813T160000
DTEND;TZID=Asia/Kolkata:20250813T170000
DTSTAMP:20250731T044838Z
URL:https://cds.iisc.ac.in/events/cds-kiac-seminar-cds-102-13th-august-tow
 ards-measuring-and-mitigating-hallucinations-in-generative-image-super-res
 olution/
SUMMARY:CDS-KIAC {Seminar}@ CDS: #102: 13th August: "Towards Measuring and 
 Mitigating Hallucinations in Generative Image Super-Resolution"
DESCRIPTION:We welcome you to CDS-KIAC talk on 13th August 2025 (Wednesday)
 . The details are as below:\n\n\n\nSpeaker: Raghav Goyal\, Researcher at S
 amsung AI-Center Toronto.\nTitle: Towards Measuring and Mitigating Halluci
 nations in Generative Image Super-Resolution\nDate and Time: August 13\, 2
 025: 04:00 PM\nVenue: #102\, CDS Seminar Hall.\n\n\n\nAbstract: Generativ
 e super-resolution (GSR) currently sets the state-of-the-art in terms of p
 erceptual image quality\, overcoming the "regression-to-the-mean" blur of 
 prior non-generative models. However\, from a human perspective\, such mod
 els do not fully conform to the optimal balance between quality and fideli
 ty. Instead\, a different class of artifacts\, in which generated details 
 fail to perceptually match the low resolution image (LRI) or ground-truth 
 image (GTI)\, is a critical but under-studied issue in GSR\, limiting its 
 practical deployments. In this talk\, I will focus on measuring\, analyzin
 g\, and mitigating these artifacts (i.e.\, "hallucinations"). First\, we a
 nalyse hallucinations by observing that they are not well-characterized wi
 th existing image metrics or quality models\, as they are orthogonal to bo
 th exact fidelity and no-reference quality. Second\, to measure hallucinat
 ions\, we propose to take advantage of a multimodal large language model (
 MLLM) that assesses hallucinatory visual elements and generates a "Halluci
 nation Score" (HS) which is closely aligned with human evaluations. Third\
 , to mitigate hallucinations\, we find that certain deep feature distances
  have strong correlations with HS\, and therefore we propose to align the 
 GSR models by using such features as differentiable reward functions to mi
 tigate hallucinations.\n\nBio of Speaker: Raghav Goyal is a researcher at
  Samsung AI-Center Toronto. He obtained his PhD at University of British C
 olumbia (UBC) supervised by Prof. Leonid Sigal with a focus on data-effici
 ent learning for structured vision tasks. Prior to this\, he spent three y
 ears at a startup named “20bn” (now at Qualcomm Research) working on v
 ideo understanding including Something-Something dataset. He has published
  in top venues such as CVPR\, ICCV and NeurIPS\, with internships at Googl
 e\, Meta\, and Xerox Research. He obtained an Integrated M.Tech. (5-year p
 rogramme) from Indian Institute of Technology (IIT) Delhi in Mathematics a
 nd Computing.\n\nHost Faculty: Prof. Venkatesh Babu\n\n\n\nALL ARE WELCOME
CATEGORIES:Events,Talks
END:VEVENT
BEGIN:VTIMEZONE
TZID:Asia/Kolkata
X-LIC-LOCATION:Asia/Kolkata
BEGIN:STANDARD
DTSTART:20240813T160000
TZOFFSETFROM:+0530
TZOFFSETTO:+0530
TZNAME:IST
END:STANDARD
END:VTIMEZONE
END:VCALENDAR