A Comparison of 5 Layer and 8 Layer CNN Data Augmentation for Facial Emotion Recognition

Among the basic facial emotions that academics around the world look for are anger, disgust, fear, happiness, neutrality, sadness, and surprise. Because human interactions and many procedures have gone digital in the modern era, it is critical to detect these emotions. Using deep learning techniques like a Convolutional Neural Network (CNN), this study examines emotion detection algorithms. Training and testing were conducted using a few additional private photographs and the FER2013 benchmark dataset for facial emotion identification. Through the use of data augmentation techniques, such as image flipping, rotation, brightness correction, and contrast enhancement, this study sought to compare and evaluate the effectiveness of the two CNN 5Layer and 8Layer methods. Two CNN architectures were implemented and trained using a custom constructed dataset extracted from the FER2013 database using the Keras deep learning toolkit. Experimental results showed that both models achieved much greater test accuracy when trained using the proposed augmentation procedures, demonstrating the advantage of data augmentation in boosting emotion detection performance.