BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//132.216.98.100//NONSGML kigkonsult.se iCalcreator 2.20.4//
BEGIN:VEVENT
UID:20260411T035210EDT-3564bfzVUd@132.216.98.100
DTSTAMP:20260411T075210Z
DESCRIPTION:Abstract\n\n \n\nExisting state-of-the-art recognition models a
 chieve impressive performance but require a complete scene which may not a
 lways be available. For example\, sensing a complete scene at once is infe
 asible in applications such as aerial imaging. Further\, in applications s
 uch as disaster recovery\, imaging devices should be light\, inexpensive\,
  and energy-efficient\; thus\, they are often built using small field-of-v
 iew cameras that capture only a part of a scene at a time. In the above ca
 ses\, the imaging devices must scan the area sequentially. Moreover\, they
  must also prioritize the scanning of informative subregions for timely re
 cognition.\n\nMany developed attention models that recognize a scene by ob
 serving it through small informative subregions called glimpses. However\,
  most models locate informative glimpses by glancing at a low-resolution g
 ist of a complete scene\, which is unavailable in practice. In this thesis
 \, we develop sequential recognition models that locate and attend to info
 rmative glimpses without assessing a complete scene. Our sequential attent
 ion models predict the location of the next glimpse based solely on past g
 limpses. Our models achieve effective attention policies under partial obs
 ervability by selecting subsequent glimpses that\, combined with past glim
 pses\, help the most in reasoning about the complete scene.\n\nWe present 
 three attention models\, two for spatial and one for spatiotemporal recogn
 ition. The first is Probabilistic Attention Model (PAM). PAM uses Bayesian
  Optimal Experiment Design to attend to a glimpse with maximum expected in
 formation gain (EIG). It synthesizes features of the complete scene from p
 ast glimpses to estimate the EIG for yet unobserved regions. The second is
  Sequential Transformers Attention Model (STAM)\, which employs the one-st
 ep actor-critic algorithm to attend to a sequence of glimpses that produce
  class distribution consistent with the one produced using a complete scen
 e. The third is Glimpse Transformer (GliTr). GliTr learns an effective att
 ention mechanism for online action recognition by selecting glimpses with 
 features and class distribution consistent with the corresponding complete
  video frames.\n\nThroughout the thesis\, we evaluate our models on multip
 le datasets and compare them with existing models. Our two key findings ar
 e as follows. First\, reasoning about the complete scene from partial obse
 rvations helps in learning an effective attention policy under partial obs
 ervability. Second\, while reducing the amount of sensing required for rec
 ognition\, our glimpse-based models achieve comparable or higher performan
 ce than the existing models that require complete scenes. The key takeaway
  is that one can attain good performance even using low-cost sensing devic
 es and non-ideal imaging by automating the sensing process and compelling 
 the recognition model to fill in the missing information.\n
DTSTART:20230613T180000Z
DTEND:20230613T200000Z
LOCATION:Room 603\, McConnell Engineering Building\, CA\, QC\, Montreal\, H
 3A 0E9\, 3480 rue University
SUMMARY:PhD defence of Samrudhdhi Rangrej – Visual Hard Attention Models Un
 der Partial Observability
URL:https://www.mcgill.ca/ece/channels/event/phd-defence-samrudhdhi-rangrej
 -visual-hard-attention-models-under-partial-observability-348659
END:VEVENT
END:VCALENDAR