Refining Deep Generative Models via Discriminator Gradient Flow

Deep generative modeling has seen impressive advances in recent years, to the point where it is now commonplace to see simulated samples (e.g., images) that closely resemble real-world data. However, generation quality is generally inconsistent for any given model and can vary dramatically between samples... (read more)

PDF Abstract ICLR 2021 PDF ICLR 2021 Abstract

Results from the Paper


 Ranked #1 on Image Generation on CIFAR-10 (Frechet Inception Distance metric)

     Get a GitHub badge
TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK RESULT BENCHMARK
Image Generation CIFAR-10 SNGAN + DGflow Inception score 9.35 # 9
Frechet Inception Distance 9.62 # 1
Text Generation One Billion Word WGANGP + DGflow JS-4 0.186 # 1

Methods used in the Paper


METHOD TYPE
GAN
Generative Models
Normalizing Flows
Distribution Approximation