ReSyn
收藏资源简介:
ReSyn数据集是由上海交通大学和腾讯Youtu Lab联合创建的一个大规模图像恢复数据集,旨在解决现有数据集在图像复杂度分布上的不平衡问题。该数据集包含12000张图像,涵盖了从0.25K到4K的多种分辨率,其中30%为高质量的合成图像。数据集的创建过程中,采用了基于灰度共生矩阵(GLCM)的图像复杂度评估方法进行筛选,以确保图像复杂度的平衡分布。ReSyn数据集主要应用于图像恢复任务,旨在提升图像恢复模型的训练效果和泛化能力。
The ReSyn Dataset is a large-scale image restoration dataset jointly developed by Shanghai Jiao Tong University and Tencent Youtu Lab. It is designed to solve the problem of imbalanced image complexity distribution in existing datasets. The dataset contains 12,000 images with resolutions ranging from 0.25K to 4K, 30% of which are high-quality synthetic images. During the dataset construction phase, an image complexity evaluation method based on gray level co-occurrence matrix (GLCM) was adopted for screening, ensuring a balanced distribution of image complexity across the dataset. The ReSyn Dataset is primarily applied to image restoration tasks, with the aim of improving the training performance and generalization capability of image restoration models.




