ProceduralImageProgramsforRepresentation Learning-SupplementaryMaterial

Ascanbeseen,MixUpimprovesFIDbyabigmargin for S-21k (around 6 points), and outperforms the FID for other datasets. After training StyleGAN v2 with Shaders-21k, we produce interpolated images by sampling from the networkusing truncation. S-3.1 VTABdetailedresults In Table 3, 4 and 5 we show detailed results per dataset on the VTAB benchmark.