BEGIN:VCALENDAR
VERSION:2.0
PRODID:icalendar-ruby
CALSCALE:GREGORIAN
X-WR-CALNAME:PhD Oral Preliminary Examination – Peng Lei
X-WR-TIMEZONE:Pacific Time (US & Canada)
BEGIN:VEVENT
DTSTAMP:20260605T231310Z
UID:tag:localist.com\,2008:EventInstance_3661572
DTSTART:20180607T170000Z
DTEND:20180607T190000Z
DESCRIPTION:Pixel- and Frame-level Video Labeling using Spatial and Tempora
 l Convolutional Networks\n\nThis research report addresses the problem of 
 video labeling\, at the frame and pixel levels\, using deep learning. For 
 semantic pixel labeling\, in our initial work\,  we have developed recurre
 nt temporal deep field (RTDF). RTDF is a conditional random field (CRF) th
 at combines a deconvolution neural network and a recurrent temporal restri
 cted Boltzmann machine (RTRBM)\, which can be jointly trained end-to-end. 
 We have derived a mean-field inference algorithm to jointly predict all la
 tent variables in both RTRBM and CRF. Also\, our previous work on pixel la
 beling has addressed boundary flow estimation using a fully convolutional 
 Siamese network (FCSN). The FCSN first estimates object boundaries in two 
 consecutive frames and then predicts boundary correspondences in the two f
 rames. For frame labeling\, we have specified a temporal deformable residu
 al network (TDRN)\, which computes two parallel temporal streams: i) Resid
 ual stream that analyzes video information at its full temporal resolution
 \, and ii) Pooling/unpooling stream that captures long-range visual cues. 
 The former facilitates local\, fine-scale action segmentation\, and the la
 tter uses multiscale context for improving the accuracy of frame classific
 ation. Leveraging our previous work\, we propose two related lines of rese
 arch. The first study will introduce new regularizations in learning of th
 e temporal convolutional network (TCN) aimed at extracting meaningful temp
 oral patterns and their relevance scores for frame-level labeling. The sec
 ond study will guide a 3D convolutional (C3D) segmentation network for pix
 el-level labeling using only action- or activity-level labels as supervisi
 on.\n\nMajor Advisor: Sinisa Todorovic\nCommittee: Fuxin Li\nCommittee: Xi
 aoli Fern\nCommittee: Raviv Raich\nGCR: Leonard Coop
GEO:44.567164;-123.278692
LOCATION:Kelley Engineering Center\, 1007
SUMMARY:PhD Oral Preliminary Examination – Peng Lei
URL;VALUE=URI:https://events.oregonstate.edu/event/phd_oral_preliminary_exa
 mination_peng_lei
CATEGORIES:Lecture or Presentation
END:VEVENT
END:VCALENDAR