Unmasking The Artist: Discriminating Human-Drawn And AI-Generated Human Face Art Through Facial Feature Analysis
Minh‐Quang Nguyen, Khanh-Duy Ho, Hoang-Minh Nguyen, Canh-Minh Tu, Minh–Triet Tran, Trong-Le Do · 2023
In the 1990s, the world witnessed a revolutionary breakthrough in the realm of AI-generated art, where its applications surpassed mere visual effects. An ever-increasing number of AI-generating applications emerged, posing a critical challenge for many art contest organizations. These organizations stipulate that the artwork presented must result from human creativity without any help from AI-generated applications. This dilemma accentuates the urgency of drawing a clear distinction between AI-generated and human-made artwork, a task of paramount importance. After observing multiple works of art from AI and humans, the authors notice that the Gradient-Based feature can be utilized to differentiate images generated from different sources of generators. Therefore, the authors decided to create the dataset by collecting paintings from a state-of-the-art AI-generating art model, using algorithms for extracting features with multiple models for detecting AI-generated images and evaluating their performance. Solving this problem can help many art contest organizations as they require artists to create artwork rather than AI.