Natural vs Non-Natural Facial Expression Detection Performance Comparison #1393852 (License: Personal Use)

The chart displays an asymmetric measure based on a quantity index, evaluating how well CK+, FG-Net, and LTI-HIT models distinguish natural from non-natural facial expressions across two datasets: DA and SOM. CK+ achieves the highest performance on SOM (≈0.65), while FG-Net excels on DA (≈0.58); LTI-HIT consistently performs moderately across both. The y-axis represents normalized percentage values ranging from 0.000 to 0.700.

Used in academic or technical web pages discussing facial expression recognition benchmarks; supports research comparisons, model evaluation sections, or AI/ML conference presentations targeting computer vision practitioners.