Motion Coherent Tracking with Multi-label MRF optimization

(This work won the best student paper award in BMVC 2010)

Two successfully tracked long sequences.

David Tsai
Matthew Flagg
James M.Rehg
We present a novel off-line algorithm for target segmentation and tracking in video.
In our approach, video data is represented by a multi-label Markov Random Field model,
and segmentation is accomplished by finding the minimum energy label assignment. We
propose a novel energy formulation which incorporates both segmentation and motion
estimation in a single framework. Our energy functions enforce motion coherence both
within and across frames. We utilize state-of-the-art methods to efficiently optimize over
a large number of discrete labels. In addition, we introduce a new ground-truth dataset,
called SegTrack, for the evaluation of segmentation accuracy in video tracking. We compare
our method with two recent on-line tracking algorithms and provide quantitative and
qualitative performance comparisons.

Download PDF (8.9 MB)

Download AVI (10.8 MB)
   author = {David Tsai and Matthew Flagg and James M.Rehg},
   title = {Motion Coherent Tracking with Multi-label MRF optimization},
   journal = {BMVC},
   year = {2010},
Download slides
SegTrack Database

SegTrack database with ground-truth can be downloaded HERE.

Note: If you downloaded this database before 02/20/2011, please do it again. The "birdfall" sequence in the original package was not the right one. Thank Yong Jae Lee for pointing this out.

Code to compute metrics can be downloaded HERE.

This research is supported by:
  • NSF Grant 0916687
  • Google Research

The documents contained in these directories are included by the contributing authors as a means to ensure timely dissemination of scholarly and technical work on a non-commercial basis. Copyright and all rights therein are maintained by the authors or by other copyright holders, notwithstanding that they have offered their works here electronically. It is understood that all persons copying this information will adhere to the terms and constraints invoked by each author's copyright. These works may not be reposted without explicit permission of the copyright holder.