The Harvard Robotics Laboratory

Profs. Roger Brockett, David Mumford, Woody Yang, and Alan Yuille. The group consists of four Professors, four postdocs and twenty graduate students. It works on topics in vision, robotics, VLSI design and neural networks. The vision work includes object recognition, image segmenation, and depth estimation. (See Web page for more details).


The Robotics Lab is based on rooms 311-313 in Cruft. Cruft is connected to Pierce Hall, 29 Oxford Street.

To reach the robotics lab from MIT:

(I) Via subway . Take the Red line to Harvard Square. Enter Harvard Yard and proceed through to the Science Center [a modern building looking like an upside down polaroid camera] (about three minutes walk). Walk through the Science Center and continue along the pathway past the Music Building (on left) and a large white tank (on the right). The first building you come to will be Jefferson and continuing to the right will take you to Cruft (the next building will be Pierce Hall).

(II) Via car . From the Cambridge Marriott, take Broadway west right to the end. Then bear right onto Quincy Street. Turn left at the traffic lights onto Kirkland, then right at Oxford Street (only option). Cruft will be on your left a couple of hundred yards down. Cruft is next to Pierce Hall (29 Oxford Street), a large red brick building with a grass lawn.

Demonstration Abstracts

Various posters and videos will be on display. In addition there will be demonstrations on:

Song Chun Zhu
A FLEXIBLE OBJECT RECOGNITION SYSTEM for recognizing animate objects from silhouettes.

Face recognition under extreme lighting conditions
Peter Hallinan
This includes geometrical eigenwarps of images and linear combination of eigenimages to model extreme illumination changes.

Robust Medial Axis estimation
Robert Ogniewicz
This system uses Voronoi diagrams and heat diffusion to robustly compute the medial axes of complex shapes. Code is available.

Region Competition
Song Chun Zhu
A fast general purpose algorithm for region segmentation.

Tactile recognition using a CCD camera inside the figure which images the inside of the fingertip
Nicola Ferrier
The fingertip is constructed of rubber membrane and has an pattern on it which is deformed by contact with an object. These deformations are sensed by the camera and the shape of the object can be estimated.

A real time face recognition system with a VLSI implementation
Woody Yang