This paper presents a computational approach to automatically infer the aesthetic quality of photographs using their visual content. The study uses a peer-rated online photo sharing community, Photo.net, as a data source. The researchers extracted 56 visual features based on the intuition that they can distinguish between aesthetically pleasing and displeasing images. These features include exposure, colorfulness, saturation, hue, the rule of thirds, familiarity, wavelet-based texture, size and aspect ratio, region composition, low depth of field indicators, and shape convexity. Automated classifiers were built using support vector machines and classification trees, while linear regression on polynomial terms of the features was also applied to infer numerical aesthetics ratings. The study found a strong correlation between aesthetics and originality ratings, and decided to focus on aesthetics ratings only, as they are more clearly defined and less influenced by semantics. The researchers developed a classifier to qualitatively distinguish between high and low aesthetic value images and a regression model to quantitatively predict aesthetics scores. The results showed that the SVM-based classifier achieved a high accuracy of 70.12% using 15 features, with precision of 68.08% for detecting high class and 72.31% for low class. The regression model achieved a residual sum-of-squares error of 0.5020, indicating that visual features can predict human-rated aesthetics scores with some success. The study also found that certain visual properties, such as low depth of field indicators, colorfulness, and shape convexity, are important in determining aesthetic quality. The results suggest that computational approaches can help understand the relationship between human emotions and the visual content of photographs. Future work includes incorporating new features like dominant lines, converging lines, light source classification, and subject-background relationships to improve accuracy.This paper presents a computational approach to automatically infer the aesthetic quality of photographs using their visual content. The study uses a peer-rated online photo sharing community, Photo.net, as a data source. The researchers extracted 56 visual features based on the intuition that they can distinguish between aesthetically pleasing and displeasing images. These features include exposure, colorfulness, saturation, hue, the rule of thirds, familiarity, wavelet-based texture, size and aspect ratio, region composition, low depth of field indicators, and shape convexity. Automated classifiers were built using support vector machines and classification trees, while linear regression on polynomial terms of the features was also applied to infer numerical aesthetics ratings. The study found a strong correlation between aesthetics and originality ratings, and decided to focus on aesthetics ratings only, as they are more clearly defined and less influenced by semantics. The researchers developed a classifier to qualitatively distinguish between high and low aesthetic value images and a regression model to quantitatively predict aesthetics scores. The results showed that the SVM-based classifier achieved a high accuracy of 70.12% using 15 features, with precision of 68.08% for detecting high class and 72.31% for low class. The regression model achieved a residual sum-of-squares error of 0.5020, indicating that visual features can predict human-rated aesthetics scores with some success. The study also found that certain visual properties, such as low depth of field indicators, colorfulness, and shape convexity, are important in determining aesthetic quality. The results suggest that computational approaches can help understand the relationship between human emotions and the visual content of photographs. Future work includes incorporating new features like dominant lines, converging lines, light source classification, and subject-background relationships to improve accuracy.