Progress in Automatic Giraffe Recognition

The Forsyth / Fleck ‘Human Nude Detector’ system attracted widespread media attention late last century – featuring, for example, in the pages of Wired magazine. (Paper for Reference: Forsyth, D.A. and Fleck, M. M., “Automatic Detection of Human Nudes,” International Journal of Computer Vision , 32 , 1, 63-77, August, 1999). Less well publicised though were the subsequent developments for naked mammal detection  (also featuring the work of professor David Forsyth).
viz. The implementation a system which could not only detect unclothed tigers and zebras in photographs and videos, but also giraffes.

“Building a kinematic model for a giraffe is difficult (even given a collection of giraffe pictures with the background masked out) because a giraffe deforms non-rigidly. Our algorithm suggests a good kinematic model is one where the segments appear in a lot of giraffe pictures. This is why one should  model the giraffe neck as three segments rather than one; some of the giraffe pictures have a deformed neck.”

Reference : Using Temporal Coherence to Build Models of Animals (Intl. Conf. on Computer Vision (ICCV), Nice, France, Oct 2003) Note: The illustration is from a 2005 follow-up paper :Ramanan, D., Forsyth, D. A., Barnard, K. “Detecting, Localizing, and Recovering Kinematics of Textured Animals.” Computer Vision and Pattern Recognition (CVPR), San Diego, CA, June 2005