CNN-RNN Deep Learning Networks for Pattern Recognition Problems

Neural networks have remarkably solved many problems. A giant network with a high processing requirement and long training time is being investigated owing to having a massive amount of data. Therefore, parallel processing is essential for practical applications of network training. This study aims to develop the best neural network to accelerate a face recognition system. The standards used to compare tasks were speed and accuracy used to develop an algorithm providing the best performance in a parallel system. Moreover, this study aimed to combine neural networks and parallel programming to develop a swift person recognition system in a camera stream. Video frames were saved, and the images were fed to a neural network built in a parallel structure. This study provides readers with a better understanding of the new and current trends of artificial neural networks (ANN) models, successfully addressing PR issues to permit research emphasis and themes. The system used libraries, such as MPI, graphical computing compute unified device architecture (CUDA) libraries, and the softmax function to provide the required performance with high accuracy and good delay results. The results of this study showed that the parallel system using a conventional neural network (CNN) was 30% better than the serial system for face recognition and that the CNN networks provided the best accuracy among all the other systems.