A Parallel Connectionist Viewpoint Quasi-Invariant Iconic Object Recognition System

In a previous MSc project at the Department of Artificial Intelligence, an iconic object recognition system was developed. The system consists of two parts: a visual attention system that locates objects of interest, and a multi-variate matching algorithm that identifies the located objects. The conclusion reached in the MSc thesis was that although the system was biologically plausible and presented a promising approach to machine vision, it was too slow to perform reliable tests. The aim of this project was to significantly reduce the execution time of the program by parallelising the code. This objective was accomplished. The average speedup of the code is five times for a normal program workload. EPCC-SS96-13 2 Contents 1 Introduction 3 2 Background 4 3 Parallel Approach 7 4 Implementation 10 5 Results 12 6 Future Work 14 7 Conclusion EPCC-SS96-13 3 1 Introduction Image recognition plays a central part in today's Artificial Intelligence research. Its applications vary from onl..