We describe a flexible object recognition and modelling system (FORMS)
which represents and recognizes animate objects from their silhouette
s. This consists of a model for generating the shapes of animate objec
ts which gives a formalism for solving the inverse problem of object r
ecognition. We model all objects at three levels of complexity: (i) th
e primitives, (ii) the mid-grained shapes, which are deformations of t
he primitives, and (iii) objects constructed by using a grammar to joi
n mid-grained shapes together. The deformations of the primitives can
be characterized by principal component analysis or modal analysis. Wh
en doing recognition the representations of these objects are obtained
in a bottom-up manner from their silhouettes by a novel method for sk
eleton extraction and part segmentation based on deformable circles. T
hese representations are then matched to a database of prototypical ob
jects to obtain a set of candidate interpretations. These interpretati
ons are verified in a top-down process. The system is demonstrated to
be stable in the presence of noise, the absence of parts, the presence
of additional parts, and considerable variations in articulation and
viewpoint. Finally, we describe how such a representation scheme can b
e automatically learnt from examples.