Aesthetic Evaluation Based on Complex Networks
Yao Wang, Minghui Wang, Chenyi Shen, Jing Yuan Wu, Shijian Luo · 2024
With the rapid advancement of AIGC, evaluating the aesthetics of intelligently generated images has become a prominent research focus. This study introduces a novel framework for aesthetic evaluation using a complex network approach. Here, each generated image serves as a node, and their relationships form the edges of the complex network. To illustrate our approach, we employed a kitten travel scenario, generating 321 images and organizing them into two datasets for analysis. Our findings reveal that the aggregation coefficients for the two networks were 1.0317 and 1.355, with corresponding average distances of 1.0317 and 1.355. Network1 exhibited a notably large aggregation coefficient and a small average distance, indicative of its small-world characteristic.