Publications

• M. Livingston, E. Herbst. "Interactive Operations for Understanding Embedded Sensor Domains". Poster, ACM Symposium on Interactive 3D Graphics and Games, 2005.
• M. Livingston, E. Herbst. "Interactive Operations for Visualization of Ad-Hoc Sensor System Domains". IEEE Conference on Mobile Ad-Hoc and Sensor Systems, 2005.

  → We looked at displaying the coverage and connectivity of sensor networks and modeling their interaction with interesting objects--for example, magnetometers and vehicles--and displaying the results of changes to network structure.

• L. Mueller, T. Solow, N. Taylor, E. Skwarecki, R. Buels, J. Binns, C. Lin, M. Wright, R. Ahrens, Y. Wang, E. Herbst, E. Keyder, N. Menda, D. Zamir, S. Tanksley. "The SOL Genomics Network: a Comparative Resource for Solanaceae Biology and Beyond". Plant Physiology, v138, 2005.

  → I worked five semesters at a bioinformatics lab at Cornell. I did some original work on efficiently training an HMM-based classifier for signal peptides in plant proteins, and a lot of web scripting.

• P. Koehn, H. Hoang, C. Callison-Burch, M. Federico, N. Bertoldi, B. Cowan, W. Shen, C. Moran, R. Zens, C. Dyer, O. Bojar, A. Constantin, E. Herbst. "Moses: Open Source Toolkit for Statistical Machine Translation". Association for Computational Linguistics, 2007.

  → A conference version of the final report of the 2006 NSF Workshop on Human Language Technologies. We wrote a phrase-based decoder from scratch. I worked on the phrase sequence model and on error analysis.

• B. Fransen, E. Herbst, A. Harrison, W. Adams, J. Trafton. "Real-Time Face and Object Tracking". IEEE International Conference on Intelligent Robots and Systems (IROS), 2009.

  → Putting a 3-d motion model and a projection model into the inverse compositional image alignment algorithm [S. Baker, I. Matthews, "Equivalence and Efficiency of Image Alignment Algorithms", IEEE Computer Vision and Pattern Recognition, 2001] and doing face tracking. Competitor to the Watson face tracking system, which uses a 2-d model as of 2007. Morency (author of Watson) has a 2008 paper that suggests he's doing 3-d tracking, but he doesn't give details. Regardless, far as we know he can't be using an algorithm any more efficient than this.

• E. Herbst, S. Seitz, S. Baker. "Occlusion Reasoning for Temporal Interpolation using Optical Flow". Department tech report # UW-CSE-09-08-01. See http://www.cs.washington.edu/research/tr/techreports.shtml.

  We present an optical-flow-based algorithm to smoothly interpolate between two images. We reason about the depth ordering of objects, and show how bidirectional flow can be used to reduce holes in the estimated flow at the interpolated time and perform occlusion reasoning. We develop a purely pixel-wise algorithm and then add spatial regularization. We evaluate our algorithm on the interpolation set of the Middlebury flow benchmark.