Fig. 3

Visualization of spatial relationships between the C-H space and the observed image diversity in DeviantArt and Bēhance data as measured by the image similarity. (a, b, c) Degrees of sample density, the IE similarity (sample mean), and the SIFT similarity (subsample mean) of the DeviantArt images per bin in the C-H region. (d, e, f) Degrees of sample density, the IE similarity (sample mean), and the SIFT similarity (subsample mean) of the Bēhance images per bin in the C-H region. Each C-H bin (C ≈ 0.01, H ≈ 0.02) per platform occupies more than 50 images.