Rapidata/human-style-preferences-images

Rapidata图像生成偏好数据集是最大的文本到图像模型人类偏好数据集之一，包含超过1,200,000个人类偏好投票。该数据集通过Rapidata的Python API在不到100小时内收集，覆盖了全球145多个国家的参与者。数据集用于比较多个先进的图像生成模型，包括Imagen-3、Flux-1.1-pro、Flux-1-pro、Dalle-3、Midjourney-5.2和Stabel-Diffusion-3。应用场景包括基准测试新图像生成模型、开发更好的生成模型评估指标、理解全球对AI生成图像的偏好等。

This is a large-scale human preference dataset for evaluating text-to-image models. The dataset contains over 1,200,000 human preference votes collected in less than 100 hours, with participants from around the globe. The prompts are carefully curated to test various aspects of image generation. The dataset is used for benchmarking, developing evaluation metrics, understanding global preferences, training and fine-tuning image generation models, and researching cross-cultural aesthetic preferences. The data collection is powered by Rapidatas innovative annotation platform, which offers speed, global coverage, quality assurance, and cost-effectiveness.