Facial Emotion Recognition using Convolution Neural Network

With the transition from laboratory-controlled to daunting in-the-wild conditions of facial Emotion recognition (FER) and the recent popularity of deep learning strategies in various fields, deep neural networks[1], [2] have rapidly been leveraged to train discriminatory representations for automated FER. The FER will allow us to recognize the Emotion of the human face that is a major blow to recent technological development. Recent FER programmers are typically concentrating on two critical issues: overfitting due to lack of appropriate training evidence and emotion-related differences such as lighting, head posture and identification bias. It encourages to improve the other innovations, such as incorporating FER into the robotic device in order to provide the robot feelings. A systematic survey of deep FER including datasets and algorithms that offer insight into these inherent issues. First, the FER scheme with the relevant context information and recommendations for the implementations to be applied at each level. Introducing the relevant datasets that are commonly used in the literature and including agreed data collection and assessment criteria for these data sets. Competitive results on commonly used metrics that can be used to develop the project to the maximum degree that, in essence, helps more developments and the environment. The main motivation to work is to improve the accuracy for the following model statement which could make an impact on the future work. Along with that a comparative analysis between the used model and transfer learning model will provide a proper novelty for the research work.