Study compares CNNs and handcrafted features for facial expression recognition

A new study empirically compares handcrafted feature learning methods like Histogram of Oriented Gradients (HOG) and Local Binary Patterns (LBP) against Convolutional Neural Networks (CNNs) for facial expression recognition. The research utilized three datasets: FER-2013, CK+, and KDEF, evaluating HOG with Support Vector Machines (SVM), LBP with Logistic Regression, and a lightweight CNN. Results indicate that CNNs generally outperform other methods, especially on complex datasets, while HOG shows strong performance in controlled settings, and LBP performs poorly across the board. The study emphasizes the critical role of dataset complexity and robust feature learning in achieving effective facial expression recognition. AI

IMPACT Highlights the continued relevance of traditional feature engineering alongside deep learning for specific computer vision tasks.

RANK_REASON The cluster contains a single academic paper detailing empirical research findings. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• CK+
• FER-2013
• Histogram of oriented gradients
• Local binary patterns
• logistic regression model
• Rallage Rangika Chethiya Bandara Galkaduwa
• support vector machine

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →