Faster-than-SIFT Object Detection

Borja Peleato and Matt Jones

peleato@stanford.edu, mkjones@cs.stanford.edu

March 14, 2009

We propose the use of generic trees for realtime object search, and improve on the classication time taken by SIFT approximately by a factor of 5. Our approach also supports very fast training, taking no longer than it takes to search for an object in one candidate image.

Section 2 provides the background and an overview of the previous related work. 

Section 3 explains in detail our proposed scheme, 

before going into the analysis and results in section 4. 

Finally, section 5 reviews the main features of our method and gives some possible directions for future work.

generic-trees.pdf