ENHANCED REAL-TIME FACIAL EMOTION DETECTION USING DEEP LEARNING AND OPEN-CV INTEGRATION

ENHANCED REAL-TIME FACIAL EMOTION DETECTION USING DEEP LEARNING AND OPEN-CV INTEGRATION

Mr. Dhananjay A. Kaushal

PG Scholar

Department of Computer Science, 

G.H. Raisoni University, Amravati, Maharashtra, India

Received on: 17 June ,2024          Revised on: 19 July ,2024          Published on: 31 July ,2024

ABSTRACT:  Real-time facial expression recognition (FER) systems have become an essential technology in an era where human-computer interaction is becoming more and more important. These technologies improve applications in a variety of industries, including healthcare, entertainment, security, and customer service, by enabling machines to comprehend and react to human emotions. This research study describes how deep learning techniques were used to design and assess a reliable real-time FER system. utilizing the FER 2013 dataset, which is accessible to the public and includes more than 35,000 tagged facial photos representing the seven emotion categories of anger, disgust, fear, happiness, neutrality, sadness, and surprise. Numerous convolutional neural network (CNN) topologies are used to train and evaluate the system.

To enhance model generalization, this research starts with data preprocessing using augmentation and normalization procedures. Next, we evaluate how well the Inception V3 network, the VGG16 architecture, and a simple CNN model recognize and classify face emotions. Our findings show that all models perform well in real-time scenarios, but the VGG16 model achieves greater accuracy and F1-scores, showing its robustness in managing the intricacies of facial expression data.

Our approach is validated by integrating the trained models into a real-time application that records live video, analyzes it frame-by-frame, and makes real-time predictions about the dominating emotion. With potential uses in emotional AI systems and adaptive user interfaces, the system exhibits excellent accuracy and reactivity. This work describes the system’s implementation and technical development as well as its consequences, difficulties, and potential future paths for real-time FER.

INDEX TERMS – Real-Time Facial Expression Recognition, Convolutional Neural Networks, Deep Learning, FER 2013 Dataset, VGG16, Inception V3, Emotion Detection, Human-Computer Interaction, Data Augmentation, Live Video Processing.

DOI LINK – https://doi.org/10.69758/GIMRJ2407II0IV12P0014

Download