BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Extending Uncertainty Quantification to New Domains in Computer Vi
 sion
DTSTART:20230628T080000
DTEND:20230628T100000
DTSTAMP:20260408T111445Z
UID:4ad63037a886b684a6986de85a9ae8c5bdb41703599a490cf3e03fb0
CATEGORIES:Conferences - Seminars
DESCRIPTION:Deniz Mercadier\nEDIC candidacy exam\nExam president: Prof. An
 toine Bosselut\nThesis advisor: Prof. Pascal Fua\nCo-examiner: Prof. Amir 
 Zamir\n\nAbstract\nDeep learning based methods have come to dominate a var
 iety of computer vision tasks with sheer predictive performance. However\,
  their use in safety critical applications remains limited due to the unre
 liability of their predictions. Uncertainty quantification methods have be
 en developed to allow these models to estimate the reliability of their ow
 n predictions. Unfortunately\, these methods are usually developed on clas
 sification and regression tasks\, and cannot be universally extended to mo
 re complex ones. One class of such tasks are those related to 3D shape mod
 eling\, applicable to a wide variety of problems in medical imaging\, manu
 facturing and simulation. Our research aims to develop tailored approaches
  for 3D tasks able to quantify uncertainty over whole surfaces in a fine-g
 rained manner\, focusing initially on explicit mesh representations.\n\nIn
  this context\, we first introduce the basics of uncertainty quantificatio
 n on classification and regression tasks\, along with contemporary challen
 ges based on an extensive survey. Next\, we examine a method to estimate u
 ncertainty for monocular depth estimation. We then discuss another recent 
 work on improving the confidence quality of object detection models. These
  complex examples finally motivate our research goal of developing methods
  for advanced 3D shape representation tasks\, an unexplored direction as o
 f yet\, to our knowledge.\n\nBackground papers\n1. A review of uncertainty
  quantification in deep learning: Techniques\, applications\, and challeng
 es\n    in Information Fusion - December 202.  Authors: Moloud Abdar\,
  et al.  Link: https://arxiv.org/pdf/2011.06225.pdf\n2. Gradient-based Un
 certainty for Monocular Depth Estimation in ECCV 2022. Authors: Julia Horn
 auer\, Vasileios Belagiannis\n    Link: https://www.ecva.net/papers/ecc
 v_2022/papers_ECCV/papers/136800598.pdf\n3. Bridging Precision and Confide
 nce: A Train-Time Loss for Calibrating Object Detection in CVPR 2023 Autho
 rs: Muhammad Akhtar Munir\,et al.\n    Link: https://openaccess.thecvf
 .com/content/CVPR2023/papers/Munir_Bridging_Precision_and_Confidence_A_Tra
 in-Time_Loss_for_Calibrating_Object_CVPR_2023_paper.pdf\n 
LOCATION:BC 229 https://plan.epfl.ch/?room==BC%20229
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR