Fr\'echet Inception Distance (FID) is the primary metric for ranking models in data-driven generative modeling. While remarkably successful, the metric is known to sometimes disagree with human judgement. We investigate a root cause of these discrepancies, and visualize what FID "looks at" in generated images. We show that the feature space that FID is (typically) computed in is so close to the ImageNet classifications that aligning the histograms of Top-$N$ classifications between sets of generated and real images can reduce FID substantially -- without actually improving the quality of results. Thus, we conclude that FID is prone to intentional or accidental distortions. As a practical example of an accidental distortion, we discuss a case where an ImageNet pre-trained FastGAN achieves a FID comparable to StyleGAN2, while being worse in terms of human evaluation.
翻译:Fr\'echet 感知距离(FID) 是数据驱动基因模型中排名模型的主要衡量标准。 虽然该指标非常成功, 但已知有时与人类判断不相符。 我们调查了这些差异的根源, 并想象FID在生成图像中“ 外观 ” 。 我们显示, FID(通常) 所计算的特征空间非常接近图像网络分类, 使生成的和真实图像组之间的最高一美元分类直方图能够大幅降低FID值, 而不会实际改善结果的质量。 因此, 我们得出结论, FID容易发生有意或意外的扭曲。 作为意外扭曲的一个实例, 我们讨论了一个案例, 一个图像网络预先训练的FastGAN 取得了与StyleGAN2相似的FID, 而在人类评估方面则更糟。