VG-RAM Weightless Neural Networks for Face Recognition
Alberto Ferreira De Souza, Claudine Santos Badue, Felipe Thomaz Pedroni, Stiven Schwanz, Hallysson Oliveira, Soterio Ferreira de Souz · InTech eBooks · 2010
In this work, we presented an experimental evaluation of the performance of Virtual Generalizing Random Access Memory Weightless Neural Networks (VG-RAM WNN Aleksander (1998)) on face recognition. We presented two VG-RAM WNN face recognition architectures, one holistic and the other feature-based, and examined its performance with two well known face database: the AR Face Database and the Extended Yale Face Database B. The AR Face Database is challenging for face recognition systems because it has images with different facial expressions, occlusions, and varying illumination conditions. The best performing architecture (feature-based) showed robustness in all image conditions and better performance than many other techniques from literature, even when trained with a single sample per person. In future works, we will examine the performance of VG-RAM WNN with other databases and use it to tackle other problems associated with face recognition systems, such as face detection, face alignment, face recognition in video, etc.