Synthetic Overhead Imagery Dataset
收藏arXiv2023-05-10 更新2024-06-21 收录
下载链接：
https://stresearch.github.io/synthetic-overhead-dataset/
下载链接
链接失效反馈官方服务：
资源简介：
Synthetic Overhead Imagery Dataset是由德雷塞尔大学创建的，包含真实、全合成和部分操纵的空中图像。数据集包含9011条图像，来源于MapBox和Google Maps的卫星图像，以及通过定制扩散模型生成的图像。创建过程中使用了多种生成策略，包括无条件和条件生成。该数据集主要用于开发和评估法医工具，特别是用于检测和定位图像中的合成内容，解决图像操纵和伪造的问题。
Synthetic Overhead Imagery Dataset was developed by Drexel University. This dataset contains aerial images that are real, fully synthetic, or partially manipulated. Comprising 9011 image samples, it is sourced from satellite imagery of MapBox and Google Maps, alongside images generated using a custom diffusion model. Multiple generation strategies were adopted during its construction, including unconditional and conditional generation. The primary purpose of this dataset is to develop and evaluate forensic tools, particularly for detecting and localizing synthetic content in images, thereby addressing the challenges of image manipulation and forgery.
提供机构：
德雷塞尔大学
创建时间：
2023-05-10
