Logo Recognition by Recursive Neural Networks

In this paper we propose an adaptive model, referred to as Recursive Neural Networks (RRNNs) for logo recognition by explicitly conveying logo item intom-ary tree representation, where symbolic and sub-symbolic information coexist. Each node in the contourtree is associated with an exterior or interior contour extracted from the logo instance. A feature vector, which includes the perimeter of the contour, the area surrounded and the number of critical points at some pre-determined intervals, is associated with each node. The pattern representation transformed in this way contains the topological structured information of logo and continuous values pertaining to each contour node. Afterwards, the RRNNs are used to learn the logo regularities expressed by contour-trees. The experimental results are reported on 40 real logos with very promising results.