BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:Turkey BEGIN:DAYLIGHT DTSTART:20380119T061407 TZOFFSETFROM:+0300 TZOFFSETTO:+0300 RRULE:FREQ=YEARLY;BYDAY=3TU;BYMONTH=1 TZNAME:+03 END:DAYLIGHT BEGIN:STANDARD DTSTART:20160907T000000 TZOFFSETFROM:+0300 TZOFFSETTO:+0300 RRULE:FREQ=YEARLY;BYDAY=1WE;BYMONTH=9 TZNAME:+03 END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20181107T141634Z UID:074AAD8C-F03A-4D0D-8187-AAD53D974F96 DTSTART;TZID=Turkey:20181102T133000 DTEND;TZID=Turkey:20181102T153000 DESCRIPTION:Speaker: Asst. Prof. Nazlı İkizler Çinbiş\, Hacettepe Unive rsity\n\nTopic: "Recognition of Human Interactions and Collective Activiti es"\n\nLocation: Middle East Technical University\, Ankara\, Turkey\n\nAbs tract: In this talk\, I will give an overview about the recent algorithms that we have been working on at the Hacettepe Computer Vision Lab regardin g human activity understanding. First\, I will introduce "Histogram of Seq uences"\, a video representation that aims at effective recognition of hum an interactions. This method basically combines the power of discriminativ e sequence mining and histogram representation\, by mining sequences of th e visual features that occur consequently in space and time. As an extensi on to base model\, we incorporate a hierarchical temporal pyramid mechanis ation\, where the height of the temporal pyramid determines the temporal s cales of the sequences. This new representation is likely to cover more co mplex spatiotemporal relationships such that the temporal variability of l ocal sequences can be more accurately modelled. In the second part of my t alk\, I will present our work on collective activity recognition that anal yses the behaviour of groups of people in videos. In this context\, I will introduce "region-based multi-stream convolutional neural networks"\, whe re we extend the successful two-stream convolutional neural network archit ecture to handle multiple regions of interest in conjunction with the regu lar RGB and optical flow streams. In this context\, we explore several way s of fusing multiple spatial and temporal streams so that the accuracy of recognition can be improved.\n\nBio: Dr. Nazlı İkizler Cinbiş received her BSc and MSc degrees from Department of Computer Engineering at Bilkent University. During 2005-2006\, she was a visiting scholar at University o f Illinois at Urbana-Champaign (UIUC). After receiving her PhD degree from Bilkent University in 2008\, she worked as a post-doctoral research assoc iate at Boston University (USA). Since 2011\, she works as an Assistant Pr ofessor at Hacettepe University Department of Computer Engineering. She is amongst the founders of the Hacettepe University Computer Vision Laborato ry (HUCVL). She is an associate editor of the IET Computer Vision journal since 2016 and has served as an area chair for CVPR 2018. Her research are as are mainly computer vision and machine learning\, specifically focusing on video processing\, human action and interaction recognition in images and videos and zero-shot learning. For more information\, please visit htt p://web.cs.hacettepe.edu.tr/~nazli\n\nSpeaker(s): Asst. Prof. Nazli Ikizle r Cinbis\, \n\nAnkara\, Ankara\, Türkiye LOCATION:Ankara\, Ankara\, Türkiye ORGANIZER:ozergul@metu.edu.tr SEQUENCE:2 SUMMARY:IEEE AP/MTT/EMC/ED TURKEY CHAPTER SEMINAR SERIES -- SEMINAR 37 URL;VALUE=URI:https://events.vtools.ieee.org/m/181388 X-ALT-DESC:Description: <br /><p>Speaker: Asst. Prof. Nazlı İkizler &Cced il\;inbiş\, Hacettepe University</p>\n<p>Topic: "Recognition of Human Int eractions and Collective Activities"</p>\n<p>Location:&nbsp\;Middle East T echnical University\, Ankara\, Turkey</p>\n<p>Abstract:&nbsp\;In this talk \, I will give an overview about the recent algorithms that we have been w orking on at the Hacettepe Computer Vision Lab regarding human activity un derstanding. First\, I will introduce "Histogram of Sequences"\, a video r epresentation that aims at effective recognition of human interactions. Th is method basically combines the power of discriminative sequence mining a nd histogram representation\, by mining sequences of the visual features t hat occur consequently in space and time. As an extension to base model\, we incorporate a hierarchical temporal pyramid mechanisation\, where the h eight of the temporal pyramid determines the temporal scales of the sequen ces. This new representation is likely to cover more complex spatiotempora l relationships such that the temporal variability of local sequences can be more accurately modelled. In the second part of my talk\, I will presen t our work on collective activity recognition that analyses the behaviour of groups of people in videos. In this context\, I will introduce "region- based multi-stream convolutional neural networks"\, where we extend the su ccessful two-stream convolutional neural network architecture to handle mu ltiple regions of interest in conjunction with the regular RGB and optical flow streams. In this context\, we explore several ways of fusing multipl e spatial and temporal streams so that the accuracy of recognition can be improved.</p>\n<p>Bio:&nbsp\;Dr. Nazlı İkizler Cinbiş received her BSc and MSc degrees from Department of Computer Engineering at Bilkent Univers ity. During 2005-2006\, she was a visiting scholar at University of Illino is at Urbana-Champaign (UIUC). After receiving her PhD degree from Bilkent University in 2008\, she worked as a post-doctoral research associate at Boston University (USA). Since 2011\, she works as an Assistant Professor at Hacettepe University Department of Computer Engineering. She is amongst the founders of the Hacettepe University Computer Vision Laboratory (HUCV L). She is an associate editor of the IET Computer Vision journal since 20 16 and has served as an area chair for CVPR 2018. Her research areas are m ainly computer vision and machine learning\, specifically focusing on vide o processing\, human action and interaction recognition in images and vide os and zero-shot learning. For more&nbsp\;<span class="s1">information\, p lease visit&nbsp\;</span><span class="s2">http://web.cs.hacettepe.edu.tr/~ nazli</span></p> END:VEVENT END:VCALENDAR