Finding an appropriate set of features is an essential problem in the desig
n of shape recognition systems. This paper attempts to show that for recogn
izing simple objects with high shape variability such as handwritten charac
ters, it is possible, and even advantageous, to feed the system directly wi
th minimally processed images and to rely on learning to extract the right
set of features. Convolutional Neural Networks are shown to be particularly
well suited to this task. We also show that these networks can be used to
recognize multiple objects without requiring explicit segmentation of the o
bjects from their surrounding. The second part of the paper presents the Gr
aph Transformer Network model which extends the applicability of gradient-b
ased learning to systems that use graphs to represents features, objects, a
nd their combinations.