Prediction of Best Optimizer for Facial Expression Detection using Convolutional Neural Network

Deep learning is some sort of machine learning which resembles the features of the human's brain while analyzing data and generating patterns in order to draw conclusions. It is a part of machine learning, which in turns a part of artificial intelligence, and it consists of networks that are capable of learning from unsupervised, unstructured, or unlabeled data. When the size of the input data is vast, deep learning can outperform machine learning. Despite traditional programs, which develop insights with data in a linear fashion, deep learning systems include a hierarchical characteristic which permits machines to analyze the data in a non-linear fashion. Deep learning is advised for complicated tasks involving massive quantities of data, such as picture classification, natural language processing, and speech recognition In the scientific world, facial expression recognition is a passionately disputed issue. Human express their feelings using facial expressions. Humans can easily distinguish facial expressions, but computers find it challenging to do so because human facial expressions might fluctuate in photos due to elements such as brightness, contrast, and so on. This research uses seven main human emotions to train computers. Sadness, contempt, fear, happiness, anger, surprise, and neutral are among the emotions. Convolutional Neural Networks have the potential to solve among the most popular complex challenges in the fields such as deep learning and computer vision: facial emotion recognition. The system is being developed in three stages: image pre-processing, detection, and classification. Every iteration, the accuracy will be enhanced by fine-tuning key parameters for example, the initial thick layer's unit count, the rate of dropout in the dropout layer, as well as optimization strategies. Various optimizers, including as Adam, Adagrad, Stochastic Gradient Descent, and Nadam, were utilized to achieve the best accuracy.