CNN-based Human Emotion Recognition from Facial Images in Controlled-Uncontrolled Environment

Emotions play a crucial role in shaping human thoughts, behavior, and feelings. Consequently, the task of recognizing emotions from facial images has attained significant attention because of its applications across various areas. Traditional algorithms have often proven inadequate in meeting real-time human needs, leading to the emergence of machine learning and deep learning algorithms as more effective alternatives. Existing researchers of computer vision and affective computing have primarily concentrated on enhancing the classification accuracy of benchmark datasets. In line with these advancements, the main focus of the study is to evolve a specialized Neural Network based model which is competent in accurately classifying seven distinct human facial emotions. The model is meticulously trained and rigorously tested using both laboratory-controlled with an accuracy of 98% and 53% over uncontrolled environment images without using any sort of pre-processing and also ensuring model robustness across different scenarios.