NÜWA: Visual Synthesis Pre-training for Neural visUal World creAtion

Benchmark Model Rank Results
text-to-image-generation-on-cocoXMC-GAN (256 x 256)#20FID: 9.3Inception score: 30.5
text-to-image-generation-on-cocoNÜWA (256 x 256)#30FID: 12.9Inception score: 27.2
text-to-image-generation-on-cocoDM-GAN (256 x 256)#44FID: 26.0Inception score: 32.2
text-to-image-generation-on-cocoCogView (256 x 256)#47FID: 27.1Inception score: 18.2
text-to-image-generation-on-cocoDALL-E (256 x 256)#48FID: 27.5Inception score: 17.9
text-to-image-generation-on-cocoAttnGAN (256 x 256)#53FID: 35.2Inception score: 23.3
text-to-image-generation-on-cocoDF-GAN (256 x 256)#58Inception score: 18.7
text-to-video-generation-on-msr-vttNUWA#12CLIPSIM: 0.2439CLIP-FID: 47.68FID: 47.68
video-generation-on-bair-robot-pushingNUWA#3FVD score: 86.9Cond: 1Train: 15Pred: 15