Avoiding the Inherent Limitations in Datasets Used for Measuring Aesthetics When Using a Machine Learning Approach

Figure 5

Results obtained for the four models proposed and optimised by hyperparameterisation. On the right, the mean values for Spearman’s (top) and Pearson’s r (bottom) are shown. On the left, the distributions of all 50 independent runs for each optimum model (Spearman top and Pearson bottom) are shown with different input sizes tested using FS.

(a)

(b)

(c)

(d)