Eran Borenstein

I am currently a VP of computer vision at Videosurf, a company that applies computer vision technology to a search engine for videos, where I am leading the development of the face-detection and recognition technology. I was a Visiting scientist at the Division of Applied Mathematics Brown University Office: Room 310, 182 George St, Providence, RI 02912 Phone: (401) 863-7422 email:

Main interests

My research focuses on machine and human vision, in particular object classification and recognition, figure-ground segmentation, the interaction between top-down and bottom-up visual processes and other problems related to perceptual organization. The aim is to better understand and imitate the processes of human vision through mathematical models that account for our ability to know "what" is in an image and "where" it is.

Academic Background

• Post-doctoral Fellow, MSRI and EECS UC Berkeley.
• Ph.D. & M.Sc., Department of Computer Science and Applied Mathematics , Weizmann Institute of Science.
• B.Sc., Electrical Engineering Department, Technion - Israel Institute of Technology.

VideoSurf

Teaching

• AM41 Mathematical methods in the brain sciences (Spring 2007).
• Introduction to Vision, Open University of Israel (2003)
  

Publications

• Context, Computation, and Optimal ROC Performance in Hierarchical Models
  L. Chang, Y. Jin, W. Zhang, E. Borenstein and S. Geman
  International Journal of Computer Vision, Volume 93, Number 2
  
  
• Combined Top-Down/Bottom-Up Segmantation
  E. Borenstein, S. Ullman.
  IEEE Trans. Pattern Anal. Mach. Intell., 2008: 2109-2125
  
  
• Shape guided object segmentation
  E. Borenstein, and J. Malik
  IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2006
  
  
• Combining Top-down and Bottom-up Segmentation
  (abstract)
  E. Borenstein, E. Sharon and S. Ullman
  Proceedings IEEE workshop on Perceptual Organization in Computer Vision
  IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2004
  
  
• Learning to Segment
  (abstract)
  E. Borenstein and S. Ullman
  Springer-Verlag LNCS 3023
  European Conference on Computer Vision (ECCV), May 2004
  
  
• Class-Specific, Top-Down Segmentation
  (abstract)
  E. Borenstein and S. Ullman
  Springer-Verlag LNCS 2351
  European Conference on Computer Vision (ECCV), May 2002
  
  
• Lightness and Brightness Computations by Retinex-like Algorithms
  (abstract)
  Thesis for my M.Sc degree with Shimon Ullman
  Weizmann Institute of Science 2000
  
  

Segmentation Databases (with manual figure-ground labeling)

The horse, runners and cars databases were collected via google image and then manually labelled for testing figure-ground segmentation. For further elaboration, please refer to our related publications above (ECCV2002, ECCV2004,CVPR2004,CVPR2006).
Please feel free to download and use these database, providing you reference one of the above papers.

Horse Database

The Weizmann Horse Database consists of 328 side-view color images of horses that were also manually segmented. The images were randomly collected from the WWW to evaluate the top-down segmentation scheme as well as its combination with bottom-up processing. Note however that in our experiments we do not use RGB information and the images are given in this format just for your conveneince.
Entire database (19.5Mb) including:
Horses (RGB) (14Mb)
Horses (grey) (4Mb)
Horses (figure-ground) (1.3Mb)