Large Scale GAN Training for High Fidelity Natural Image Synthesis

ICLR 2019 Andrew BrockJeff DonahueKaren Simonyan

Despite recent progress in generative image modeling, successfully generating high-resolution, diverse samples from complex datasets such as ImageNet remains an elusive goal. To this end, we train Generative Adversarial Networks at the largest scale yet attempted, and study the instabilities specific to such scale... (read more)

PDF Abstract

Results from the Paper


TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK RESULT BENCHMARK
Image Generation CIFAR-10 BigGAN Inception score 9.22 # 4
FID 14.73 # 11
Image Generation ImageNet 128x128 BigGAN-deep FID 5.7 # 1
IS 124.5 # 1
Image Generation ImageNet 128x128 BigGAN FID 8.7 # 4
IS 98.8 # 3
Conditional Image Generation ImageNet 128x128 BigGAN-deep FID 5.7 # 2
Inception score 124.5 # 2
Conditional Image Generation ImageNet 128x128 BigGAN FID 8.7 # 5
Inception score 98.8 # 3

Methods used in the Paper