<![CDATA[This paper presents a comparison of the classification accuracy of Logistic Regression (LR) and Support Vector Machine (SVM) classifiers on facial expression classification based on image embeddings obtained from pre-trained models like VGG-16, VGG-19, and Inception V3. Facial expression classification has relevance in emotion analysis, human-computer interaction, and security. The dataset consisted of five expressions: Angry, Fear, Happy, Neutral, and Sad. Feature embeddings were extracted by using CNN models, which are said to learn spatial features, and were classified using LR and SVM. Performance metrics like accuracy, precision, recall, and F1-score were evaluated. Inception V3 topped with 89.3% accuracy on SVM, followed by VGG-19 (87.6%) and VGG-16 (85.4%). Inception V3 was best in discriminating fine-grained expressions, as confirmed through confusion matrix analysis and visualization techniques like MDS and t-SNE. In contrast to earlier works on individual models or conventional approaches, this work emphasizes the merits of fusing powerful CNNs with strong classifiers. Limitations encompass a limited dataset and just five expressions, indicating that future research should address larger, varied datasets and real-time responsiveness for enhanced system robustness.]]>
Copyrights © 2025
