Q-EVAL-100K
收藏资源简介:
Q-EVAL-100K数据集是由上海交通大学研究团队创建的,旨在评估视觉质量和对齐级别的文本到视觉内容评价数据集。该数据集包含100K个实例,涵盖了文本到图像和文本到视频模型,共有960K个针对视觉质量和对齐度的人工标注。数据集通过精心设计的实验过程和标准,确保了标注的高质量,适用于大型多模态模型的学习,以提升视觉质量和对齐度的评价能力。
The Q-EVAL-100K dataset, developed by a research team from Shanghai Jiao Tong University, is a text-to-visual content evaluation dataset intended for assessing visual quality and alignment. It consists of 100K instances covering text-to-image and text-to-video models, with a total of 960K human annotations dedicated to visual quality and alignment evaluation. The dataset adopts meticulously designed experimental protocols and standards to ensure high-quality annotations, making it suitable for training large multimodal models to enhance their ability in evaluating visual quality and alignment.

- 1Q-Eval-100K: Evaluating Visual Quality and Alignment Level for Text-to-Vision Content上海交通大学 · 2025年



