Tag
The paper introduces Adversarial Fréchet Distance (AdvFD), which adds a learnable adversarial feature space to static Fréchet losses to improve generator post-training, with real-feature whitening to stabilize optimization.