PIPAL 图像质量评价数据集
收藏资源简介:
PIPAL 全称 Perceptual Image Processing ALgorithms,是一个感知图像处理算法的 IQA 数据集。由于图像重建 (IR) 算法的快速发展,特别是一些基于 GAN 的模型的出现,现有的图像质量评估 (IQA) 方法已经无法很好地评估这些图像重建方法了。因此,香港中文大学深圳校区的团队提出了该数据集,其中 IQA 方法应当随着 IR 算法一起演进更新,并使用了 Elo 评分系统来对两两图像进行比较,更新评分。
PIPAL, whose full name is Perceptual Image Processing ALgorithms, is an IQA dataset focused on perceptual image processing algorithms. With the rapid advancement of image reconstruction (IR) algorithms, particularly the emergence of several GAN-based models, existing image quality assessment (IQA) methods can no longer adequately evaluate these image reconstruction approaches. Therefore, the team from The Chinese University of Hong Kong, Shenzhen proposed this dataset, which requires that IQA methods should evolve and update in tandem with IR algorithms, and adopted the Elo rating system to perform pairwise image comparisons and update their ratings.




