AI design for life

Researchers have developed an artificial intelligence (AI) system that evaluates the visual appeal of literary and artistic product designs by mimicking how people naturally direct their attention across an image, a step that could help designers create products that better match consumer preferences. The work was published in the International Journal of Engineering Systems Modelling and Simulation.

The researchers suggest that existing image-aesthetics systems focus too much on isolated visual features while overlooking visual saliency. Visual saliency is the idea that parts of an image are instinctively more attractive to human attention. Additionally, those earlier models ignore composition, the arrangement of visual elements such as lines, shapes, and background within an image.

The new algorithmic method combines two approaches. The first analyses edge patterns to capture the structure and balance of a design. The second uses weakly supervised learning, a machine-learning technique that learns from limited labelled examples, and an attention mechanism, which enables the model to prioritise the most important parts of an image during analysis.

Tests on two widely used image-aesthetics datasets found that the approach outperformed established deep-learning models. Having built the system on the EfficientNet architecture that also achieved a favourable balance between accuracy, speed, and computing cost.

The team suggests that the method could provide designers with measurable guidance on how changes to composition or focal points affect perceived quality. They also suggest it could help preserve cultural identity in product design by incorporating aesthetic principles rooted in different artistic traditions, rather than relying on imitation of dominant international styles.

Yu, X. (2026) ‘Forms and innovative applications of fine arts factors in the design of literary and artistic products’, Int. J. Engineering Systems Modelling and Simulation, Vol. 17, No. 7, pp.1–11.

News media may use this press release as source material, in whole or in part, provided the content is not materially misrepresented. A link back to the original article is appreciated.