Real-RSCM
收藏资源简介:
Real-RSCM数据集由中国海洋大学信息科学与工程学院创建,包含10,000张分辨率为512x512像素的图像,从中精心挑选了8,620张复制移动篡改图像和1,380张未篡改图像作为负面样本。数据集中的图像经过人工处理,包括车辆、飞机、船只、建筑物、道路和植被等显著目标被手动缩放、旋转并复制到图像中的合理位置。每个篡改图像都配对了至少14个问题和50个可能的答案,形成了173,780个图像-问题-答案三元组。该数据集针对遥感图像的复制移动篡改检测任务,提供了更加真实的评价指标,适用于土地资源监测和国防安全等领域。
The Real-RSCM dataset was created by the College of Information Science and Engineering, Ocean University of China. It contains 10,000 images with a resolution of 512×512 pixels, among which 8,620 copy-move forgery images and 1,380 untampered images were carefully selected as negative samples. The images in the dataset were manually processed: prominent objects including vehicles, aircraft, ships, buildings, roads and vegetation were manually scaled, rotated and copied to reasonable positions within the images. Each tampered image is paired with at least 14 questions and 50 possible answers, forming a total of 173,780 image-question-answer triples. Aimed at the copy-move forgery detection task for remote sensing images, this dataset provides more realistic evaluation metrics, and is applicable to fields such as land resource monitoring and national defense security.




