Facial Image expression recognition and prediction system.

Area of Science:

• Computer Science
• Artificial Intelligence
• Machine Learning

Background:

• Facial expression recognition systems are crucial for human-computer interaction.
• Developing robust models for accurate emotion detection from facial cues is an ongoing challenge.

Purpose of the Study:

• To propose and evaluate three distinct architectural models for a facial expression prediction system.
• To compare the performance of Support Vector Machines (SVM), VGG-NET Convolutional Neural Networks (CNNs), and an enhanced CNN model.
• To identify the most effective architecture for recognizing seven distinct facial expressions.

Main Methods:

• Utilized a dataset of over 35,500 facial images representing seven different expressions.
• Implemented a Support Vector Machine (SVM) for initial classification.
• Developed a Convolutional Neural Network (CNN) using the VGG-NET architecture.
• Designed an enhanced CNN model with convolutional sequential layers for improved performance.
• Preprocessed data to minimize noise and analyzed model performance using confusion matrices, loss, and accuracy metrics.

Main Results:

• The enhanced CNN model, utilizing convolutional sequential layers, showed improved accuracy and reduced loss compared to the SVM and VGG-NET CNN models.
• Performance metrics including loss and accuracy were visualized using bar graphs and scatter plots.
• Confusion matrices were employed to quantitatively assess the performance of each implemented model.

Conclusions:

• The enhanced CNN architecture proved most effective for facial expression recognition among the evaluated models.
• The study demonstrates a user-friendly approach to emotion recognition with visualized outputs for each facial image.
• Further research can build upon this enhanced model for more sophisticated emotion analysis systems.