Table 1.

Mean classification scores (±SD) for art images vs. non-art images, according to different categories of artworks.

Category	Accuracy	Precision	Recall
Western art (m = 1629)	0.939 ± 0.008	0.931 ± 0.005	0.926 ± 0.017
Chinese art (m = 210)	0.851 ± 0.012	0.354 ± 0.018	0.852 ± 0.038
Islamic art (m = 251)	0.957 ± 0.015	0.726 ± 0.083	0.964 ± 0.008
All combined (m = 2,090)	0.930 ± 0.004	0.942 ± 0.009	0.913 ± 0.010

While for Western art (JenAesthetics dataset) and Islamic art, the classification score is above 90%, it drops to 85% for Chinese art. Here, the mean precision is low, due to an overall lower richness (i.e., higher values of P_a(22)). Results are from a 5-fold cross-validation experiment. m, number of images in the dataset.