text-2-image-Rich-Human-Feedback
收藏资源简介:
该数据集名为“Rich Human Feedback for Text to Image Models”,主要用于评估AI生成的图像在风格、连贯性和提示对齐性方面的表现。数据集包含了超过150万条来自152,684个独立人类的反馈,收集过程耗时约5天。数据集的特征包括图像、提示词、单词评分、对齐评分、连贯性评分、风格评分等。数据集还提供了热图数据,用于进一步分析图像与提示词的对齐情况。数据集的用途包括文本到图像生成、文本分类、图像分类、图像到文本生成和图像分割等任务。
This dataset is named "Rich Human Feedback for Text to Image Models". It is primarily designed to evaluate the performance of AI-generated images across three key dimensions: style, coherence, and prompt alignment. The dataset contains over 1.5 million feedback entries sourced from 152,684 unique human contributors, with the entire data collection process taking approximately 5 days. Its included features are images, prompts, word-level scores, alignment scores, coherence scores, style scores, and more. Additionally, the dataset provides heatmap data to support further analysis of the alignment between generated images and their corresponding prompts. The potential applications of this dataset cover a wide range of tasks, including text-to-image generation, text classification, image classification, image-to-text generation, and image segmentation.




