RECOD Scientific Image Integrity Dataset (RSIID)
收藏资源简介:
RECOD Scientific Image Integrity Dataset (RSIID)是由坎皮纳斯大学的RECOD实验室创建的一个大型科学伪造图像基准数据集,包含39,423张图像。该数据集通过使用RECOD Scientific Image Integrity Library (RSIIL)对真实科学图像进行操作,模拟了科学图像中最常见的伪造操作,如复制、润饰和清洁。数据集的创建旨在解决科学图像完整性领域中缺乏可用数据集的问题,特别是由于数据敏感性导致的法律障碍。RSIID数据集不仅提供了丰富的像素级真实标签,还通过新的评价指标来评估复制移动伪造检测方法的性能,特别是在科学图像领域的应用。
RECOD Scientific Image Integrity Dataset (RSIID) is a large-scale benchmark dataset for scientific forged images developed by the RECOD Lab at the University of Campinas, consisting of 39,423 images. This dataset manipulates real-world scientific images via the RECOD Scientific Image Integrity Library (RSIIL) to simulate the most prevalent forgery operations in scientific imaging, including copy-move, retouching and cleaning. The creation of RSIID aims to resolve the scarcity of accessible datasets in the field of scientific image integrity, especially the legal obstacles caused by the sensitivity of relevant scientific data. The dataset not only offers rich pixel-level ground truth annotations, but also employs novel evaluation metrics to evaluate the performance of copy-move forgery detection methods, particularly for their applications in the scientific image domain.

- 1Benchmarking Scientific Image Forgery Detectors计算机科学与工程研究所,坎皮纳斯大学 · 2021年



