adrianrm/ambient-o-clip-iqa-patches-imagenet

Ambient Diffusion Omni（环境扩散全息）是一个使用低质量、合成和分布外图像来提高扩散模型质量的框架。与传统依赖高度策划的数据集的方法不同，Ambient-o在训练期间从所有可用图像中提取有价值信号，包括通常被丢弃为“低质量”的数据。此数据集卡片是基于CLIP-IQA的ImageNet质量注释。我们不仅注释了整个图像，还注释了我们论文中使用的256、128、64和32分辨率补丁的平均质量。

Ambient Diffusion Omni (Ambient-o) is a framework for using low-quality, synthetic, and out-of-distribution images to improve the quality of diffusion models. Unlike traditional approaches that rely on highly curated datasets, Ambient-o extracts valuable signal from all available images during training, including data typically discarded as low-quality. This dataset card is for CLIP-IQA-based quality annotations for ImageNet. We annotate not just the whole image, but the average patch qualities for patches of resolution 256, 128, 64, and 32 as well, as used in our paper.