Home

About

Advanced Search

Browse by Discipline

Scientific Societies

E-print Alerts

Add E-prints

E-print Network
FAQHELPSITE MAPCONTACT US


  Advanced Search  

 
Sequential Learning of Reusable Parts for Object Detection
 

Summary: Sequential Learning of Reusable Parts
for Object Detection
S. Krempp D. Geman Y. Amit
CMLA Dept. of Math. Sciences Dept. of Statistics
Ecole Normale Superieure Johns Hopkins University University of Chicago
Cachan, France 94325 Baltimore, MD 21218-2686 Chicago, IL 60637
krempp@cmla.ens-cachan.fr geman@cis.jhu.edu amit@marx.uchicago.edu
Abstract
Our long-range goal is detecting instances from a large
number of object classes in a computationally efficient man-
ner. Detectors involving a hierarchy of tests based on edges
have been used elsewhere and shown to be quite fast on-
line. However, significant further gains in efficiency - in
representation, error rates and computation - can be re-
alized if the family of detectors is constructed from com-
mon parts. Our parts are flexible, extended edge configura-
tions; they are learned, not pre-designed. In training, object
classes are presented sequentially; the objective is then to
accommodate new classes by maximally reusing parts. Ide-
ally, the number of distinct parts in the system would grow

  

Source: Amit, Yali - Departments of Computer Science & Statistics, University of Chicago

 

Collections: Computer Technologies and Information Sciences