An Application Unlocking Image Aesthetic EvaluationExpertise

The training period for developing aesthetic perception among novice photographers is typically limited to 3-6 weeks. Suchshort-term training is likely to create a significant gap between instruction and professional practice. To address this issue, weintroduce IAAS-an interpretable image aesthetics assessment system that leverages ResNet101 and Grad-CAM to processimages in a module-based manner and annotate heatmap-based attributes. This approach enables the quantification ofaesthetics through the extraction of image features. Using Kolmogorov-Arnold Networks (KAN), the system integrates attribute-level aesthetic scores into an overall aesthetic evaluation, while a Multi-modal Large Language Model (MLLM) provides naturallanguage instructional narratives. Experiments involving 150 practitioners demonstrate the effectiveness of IAAS, with anaccuracy rate of 58.24%. The findings indicate that the system can guide learners in understanding aesthetic reasoning andhelp narrow the skill gap for novice photographers.