BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Analysis by synthesis for interpretable image collection analysis 
 - Dr. Mathieu Aubry (Tenured Researcher at Ecole des Ponts ParisTech)
DTSTART:20220726T130000Z
DTEND:20220726T143000Z
UID:TALK176810@talks.cam.ac.uk
CONTACT:Gwangbin Bae
DESCRIPTION:*Abstract*\n\nI will present our recent work on analyzing the 
 content of image collections by learning a simple prototype-based model of
  images. I will start by introducing the idea and framework of Deep Transf
 ormation Invariant image analysis in the case of image clustering [1]\, wh
 ere that a simple modification of the standard K-means algorithm can lead 
 to state of the art image clustering\, while computing distances in pixel 
 space and being easy to interpret. I will then show how the idea can be ex
 tended to perform object recovery [3]\, decomposing every image in a colle
 ction into layers derived from a small set of image prototypes. This can b
 e applied to real world data\, such as collection of Instagram images\, an
 d provide models and segmentation of repeated objects. Finally\, I will ex
 plain how a similar idea can be used to perform single view reconstruction
  from a categorical image collection without any supervision.\n\n[1] Deep 
 Transformation-Invariant Clustering\, T. Monnier\, T. Groueix\, M. Aubry\,
  NeurIPS 2020\, "link":http://imagine.enpc.fr/~monniert/DTIClustering/\n\n
 [2] Unsupervised Layered Image Decomposition into Object Prototypes\, T. M
 onnier\, E. Vincent\, J. Ponce\, M. Aubry\, ICCV 2021\, "link":https://ima
 gine.enpc.fr/~monniert/DTI-Sprites/\n\n[3] Share With Thy Neighbors: Singl
 e-View Reconstruction by Cross-Instance Consistency\, T. Monnier\, M. Fish
 er\, A. Efros\, M. Aubry\, ECCV 2022\, "link":https://imagine.enpc.fr/~mon
 niert/UNICORN/\n\n*Bio*\n\nMathieu Aubry is a tenured researcher in the Im
 agine team of Ecole des Ponts ParisTech. His work is mainly focussed on Co
 mputer Vision and Deep Learning\, and their intersection with Computer Gra
 phics\, Machine Learning\, and Digital Humanities. His PhD on 3D shapes re
 presentations obtained in 2015 at ENS was co-advised by Josef Sivic (INRIA
 ) and Daniel Cremers (TUM). In 2015\, he spent a year working as a postdoc
  with Alexei Efros in UC Berkeley.\n\n*Location*\n\nThis seminar will be r
 un as a *hybrid event*. Attendees can join in person by coming to the MIL 
 Meeting Room in Baker Building (CB2 1PZ). You can also join via Zoom\, usi
 ng the provided link.\n\nGoogle Calendar for Future Seminars: https://cale
 ndar.google.com/calendar/u/0?cid=c2pjcHN0YXM2N3QyMWU3c2FqNjBqNWNiYXNAZ3Jvd
 XAuY2FsZW5kYXIuZ29vZ2xlLmNvbQ\n
LOCATION:https://zoom.us/j/6492509351?pwd=U0hoSzJ0anlhRGhzYVFmTzltNk9wZz09
  (meeting ID: 649 250 9351 / passcode: 7mu5ZJ)
END:VEVENT
END:VCALENDAR