FLD+用流模型评估图像生成质量，更准更省数据，还能快速适配新领域。
FLD+: Data-efficient Evaluation Metric for Generative Models
- 基于流模型计算图像密度，直接获得精确对数似然值
- 仅需FID1%的数据量即可稳定评估，且对退化类型响应单调
- 可轻松重训用于医疗图像等新领域，无需重新预训练大模型
我们提出一种新型图像生成质量评估指标FLD+，其基于归一化流（normalizing flows），可对任意领域的图像计算精确对数似然。与FID相比，FLD+在噪声、遮挡、扩散步数及生成模型规模等不同退化条件下表现出强单调性，评估更可靠。由于归一化流训练稳定高效，FLD+仅需两数量级更少的样本即可获得稳定结果，远低于FID对大规模真实与生成图像特征分布间弗雷歇距离的计算需求。我们进一步将归一化流应用于低维隐空间特征，显著提升计算效率，并证明其可便捷地在新领域（如医学图像）重新训练，无需依赖ImageNet预训练的InceptionNetV3等固定网络。
原文摘要 · Abstract (English)
We introduce a new metric to assess the quality of generated images that is more reliable, data-efficient, compute-efficient, and adaptable to new domains than the previous metrics, such as Fréchet Inception Distance (FID). The proposed metric is based on normalizing flows, which allows for the computation of density (exact log-likelihood) of images from any domain. Thus, unlike FID, the proposed Flow-based Likelihood Distance Plus (FLD+) metric exhibits strongly monotonic behavior with respect to different types of image degradations, including noise, occlusion, diffusion steps, and generative model size. Additionally, because normalizing flow can be trained stably and efficiently, FLD+ achieves stable results with two orders of magnitude fewer images than FID (which requires more images to reliably compute Fréchet distance between features of large samples of real and generated images). We made FLD+ computationally even more efficient by applying normalizing flows to features extracted in a lower-dimensional latent space instead of using a pre-trained network. We also show that FLD+ can easily be retrained on new domains, such as medical images, unlike the networks behind previous metrics -- such as InceptionNetV3 pre-trained on ImageNet.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。