HarmonyIQAD
收藏资源简介:
HarmonyIQAD是由上海交通大学等机构创建的首个图像和谐化质量评估数据库,包含1350张由9种不同图像和谐化算法生成的图像及其对应的人类视觉偏好评分。该数据集基于ccHarmony数据集,涵盖了多样化的前景对象和色彩风格,旨在解决现有图像质量评估方法在和谐化任务中与人类视觉偏好不一致的问题。数据集通过主观实验收集了28,350条人类评分,广泛应用于图像和谐化算法的评估与改进。
HarmonyIQAD is the first image harmony quality assessment database developed by institutions including Shanghai Jiao Tong University. It contains 1,350 images generated by 9 distinct image harmony algorithms, along with their corresponding human visual preference scores. Built upon the ccHarmony dataset, this database covers a diverse set of foreground objects and color styles, aiming to address the inconsistency between existing image quality assessment methods and human visual preferences in harmony tasks. A total of 28,350 human ratings were collected through subjective experiments, and HarmonyIQAD is widely applied for the evaluation and improvement of image harmony algorithms.

- 1HarmonyIQA: Pioneering Benchmark and Model for Image Harmonization Quality Assessment上海交通大学 · 2025年



