POCV 2014

The Ninth IEEE Computer Society Workshop on Perceptual Organization in Computer Vision

June 23, 2014 Columbus, Ohio, USA In Conjunction with IEEE CVPR 2014


Perceptual organization is the process of establishing a meaningful relational structure over raw visual data so as visual primitives arising from common physical cause are grouped together. A driving motivation behind perceptual organization research in computer vision is to deliver compact representations and to reduce the space of hypotheses for higher-level visual tasks.

Applications in artificial vision systems include figure-ground segmentation, contour completion, stereo matching, model indexing, change detection, and activity recognition. Recent advances have triggered increased focus on methods for combining low- and mid-level perceptual processing with top-down object detection and recognition systems.

Bridging this gap remains an important open problem, but one that is especially underexplored in the context of video data. There has been a long line of work on tracking, 3D reconstruction from videos, activity recognition and object detection from videos, but the concept of grouping remains poorly integrated into most matching, geometry, and classification systems.

POCV 2014 will specifically focus on perceptual organization for video, with:

  • Motion as the underlying theme
  • Recognizing humans and understanding activities as the context
  • Invited talks from human vision and computer vision communities as the format

In a departure from previous POCV workshops, the schedule this year places a heavy emphasis on invited talks. This POCV workshop is intended to be a place for discussion and development of new ideas, rather than a primary mechanism for publishing papers.

At the same time, we want to include work beyond that of the invited speakers. We encourage submission of slides or videos, with short (3-5 minute) spotlight talks being granted to accepted submissions.

The workshop schedule is now available.