EmoSense: Deep Analysis of Real-Time Facial Expression for Multi-Class Emotions

This study explores the efficacy of CNNs in classifying images into predefined categories, highlighting advancements in automated image recognition technology. The project aims to develop a robust CNN model that can accurately identify and classify images based on their content, which has significant implications for fields requiring automated image analysis. A sequential CNN architecture featuring layers of convolution, max pooling, and dropout, complemented by fully connected layers for feature learning and classification, developed and validated using TensorFlow and Keras. The model was trained and tested on a curated dataset comprising diverse images categorized into seven distinct classes, preprocessed to normalize image sizes and pixel values to enhance model training efficiency. 8L-CThe model used in the research includes 2L-CNN, 3L-CNN, and 4L-CNN architecture are used. In that, better accuracy of $62.2 \%$ for 3-layer architecture received. The CNN demonstrated high accuracy and generalizability in image classification, proving capable of handling complex pattern recognition tasks, underscoring the potential of CNNs in transforming image processing applications.