Tian Yi Huan, a Ph.D. candidate in the Department of Spatial and Cultural Design at Kookmin University's Graduate School of Techno Design (advisor: Professor Lee Sun Jung), has published a paper titled 「Aesthetic Evaluation of Visual Perception in AI-Generated Short Films via Integrated Support Vector Machine and Attention Mechanism」 in the SCIE journal IEEE Access.

IEEE Access is an international SCIE journal published by the Institute of Electrical and Electronics Engineers (IEEE). It is a globally recognized academic journal that broadly covers convergent research between design and science/technology, including artificial intelligence, computer vision, HCI, and media art.

Through this research, Tian Yi Huan proposed a new framework, ‘ZCASNet’, to quantitatively evaluate the visual perception quality of AI-generated short films. This comes amid the trend of AI technology expanding into creative domains like film and animation.

ZCASNet integrates Convolutional Neural Networks (CNN), Multi-Head Attention (MHA), Zebra Optimization Algorithm (ZOA), and Support Vector Machines (SVM) to analyze core aesthetic indicators such as visual consistency, scene plausibility, detail realism, and temporal continuity. The paper goes beyond merely measuring technical image quality, centering instead on the sensory and cognitive experience of how humans ‘perceive’ and ‘feel’ images.

To this end, four indicators were established: visual consistency (visual integration based on Gestalt principles), scene rationality (cognitive context following narrative logic), detail realism (aesthetic sensitivity to realism and immersion), and temporal continuity (the continuity principle of film editing). These four items all reflect humanities and social science concepts—such as design theory, aesthetics, and film theory—within a computer model, representing an attempt to convert human perceptual and aesthetic judgments into data for analysis.

Student Tian Yi Huan shared his thoughts: “It is meaningful to have developed the intersection of data mining and design, which I've been interested in since my master's studies, to establish an aesthetic evaluation system for AI-based creations. In the future, I wish to continue design research that harmonizes creativity and technology through studies combining design methodology and quantitative analysis.”

Professor Lee Sun Jung of Kookmin University commented, “Tian Yi Huan's research quantifies human subjective aesthetic experiences and translates them into a structure understandable by algorithms, demonstrating how design research can expand from humanistic interpretation to scientific analysis.” She added, “It holds academic significance because, while treating emotional experiences as data, it fundamentally retains human-centered perception and questions about aesthetic values.”