Assessing Food Photo Aesthetics Through Dominant Color and Color Harmony with Adaptive Smoothing Regularization

This study explores the enhancement of aesthetic evaluation for food photographs by integrating Adaptive Smoothing Regularization (ASR) with key visual features, including Dominant Color (DC) and Color Harmony (CH). Addressing the critical challenge of accurately assessing the aesthetic quality of food images a factor essential to digital marketing and consumer engagement the research employs a variety of machine learning models such as Support Vector Machine (SVM), Decision Tree (DT), Random Forest (RF), Multi-Layer Perceptron (MLP), and deep learning models like InceptionV3,trained both with and without ASR on a publicly available food image dataset. The results demonstrate a significant improvement in model performance when ASR is combined with DC and CH features. Notably, the InceptionV3+ASR model using only DC achieved the highest sensitivity in identifying aesthetically pleasing images, outperforming other models regarding recall and Fl-score. The MLP+ASR model, when utilizing both DC and CH features, showed further enhancements, though with a slight decrease in sensitivity due to the added complexity of CH. Overall, integrating ASR with carefully selected visual features proves to be a robust approach for improving the accuracy and reliability of automated aesthetic assessments.