BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//132.216.98.100//NONSGML kigkonsult.se iCalcreator 2.20.4//
BEGIN:VEVENT
UID:20260417T071755EDT-6342zpdFCE@132.216.98.100
DTSTAMP:20260417T111755Z
DESCRIPTION:Abstract\n\nVideo understanding has become a fundamental resear
 ch area in computer vision due to its wide range of applications\, includi
 ng surveillance\, healthcare\, entertainment\, and sports analytics. With 
 the advent of deep learning and the growing availability of large-scale vi
 deo data from social media\, broadcasting\, and online platforms\, remarka
 ble progress has been achieved in recognizing human actions and interactio
 ns from videos. However\, real-world environments remain challenging due t
 o dynamic motion\, visual clutter\, occlusions\, and viewpoint variations.
 \n\nThis thesis addresses the problem of Human Interaction Recognition fro
 m Videos (HIRV) under real-world conditions. The work is divided into two 
 main parts. The first part focuses on structured environments through the 
 study of sports videos\, specifically ice hockey\, as a representative and
  demanding real-world domain. Hockey broadcast videos feature fast player 
 motion\, frequent occlusions\, complex multi-person interactions\, and low
  inter-class visual variance in penalty scenes. We propose a series of ske
 leton pose–based methods for recognizing penalties and player interactions
 . These methods address several key challenges\, including limited dataset
  size\, efficient interaction recognition via custom architecture\, and ac
 tion localization in crowded scenes. In addition\, we introduce a hockey-s
 pecific pose dataset designed to evaluate and improve pose-based human int
 eraction understanding in challenging broadcast conditions.\n\nThe second 
 part of the thesis extends the study to open-world environments by investi
 gating human interactions in retail spaces\, where individuals interact no
 t only with each other but also with their surroundings. Unlike structured
  sports scenes\, retail environments involve longer and overlapping activi
 ties and complex human–object interactions influenced by environmental fac
 tors such as product placement\, stock availability\, and store layout. In
  this study\, we analyze customer behavior and decision-making processes b
 y modeling both person–person and person–object interactions over extended
  temporal sequences. In this study\, using the skeleton pose representatio
 n\, we study the customers' behavior in a retail environment and the facto
 rs affecting their decisions.\n\nOverall\, this thesis advances human inte
 raction understanding in real-world videos through the use of skeleton-bas
 ed representations\, domain-specific dataset construction\, and frameworks
  capable of reasoning about complex multi-person interactions. The propose
 d approaches demonstrate the potential of structured pose features to enha
 nce robustness\, interpretability\, and privacy in video understanding acr
 oss both structured and unconstrained environments.\n
DTSTART:20260309T170000Z
DTEND:20260309T190000Z
LOCATION:Room 603\, McConnell Engineering Building\, CA\, QC\, Montreal\, H
 3A 0E9\, 3480 rue University
SUMMARY:PhD defence of Farzaneh Askari – Skeleton-Based Human Interaction U
 nderstanding from Real-World Videos: Applications in Sports and Retail
URL:https://www.mcgill.ca/ece/channels/event/phd-defence-farzaneh-askari-sk
 eleton-based-human-interaction-understanding-real-world-videos-371469
END:VEVENT
END:VCALENDAR