Figure 4.

(a) Example images generated by DCGANs of different depths and renderings in the right-most column. The top two rows show images from networks that were trained on low-gloss textured images, and the bottom two rows show images by networks trained on high-gloss images. (b) Mean ratings from 15 human observers. The left-most triangle shows the labels that were displayed during the experiment. (c) Mean and standard deviation of gloss ratings for each network depth. (d) Mean and standard deviation of unrealness/not an object ratings per network depth.