Image dataset (Paper: Paired mini-batch training)

This dataset was used in experiments of the paper, “Paired mini-batch training: A new deep network training for image forensics and steganalysis”. All images were generated by Bossbase dataset, “Bas, P., Filler, T., Pevn`y, T., 2011. Break our steganographic system: The ins and outs of organizing boss, in: Information Hiding, Springer. pp. 59–70”.Image size – 256 x 256For forensic – 200,000 normal gray images and codes for generating manipulated images.Manipulation type – Additive white Gaussian noise deviation 1, Median filtering with a 3 x 3 kernel, Gaussian blurring with a 3 x 3 kernel and a standard deviation 0.4.For steganalysis – 160,000 normal gray images and stego images.Stego type – LSB, S-UNIWARD, WOW本数据集用于论文《配对小批量训练：一种面向图像取证（image forensics）与隐写分析（steganalysis）的新型深度网络训练方法》的相关实验。所有图像均源自Bossbase数据集（Bas, P., Filler, T., Pevný, T., 2011. Break our steganographic system: The ins and outs of organizing boss, in: Information Hiding, Springer. pp. 59–70）。 图像分辨率为256×256像素。 针对图像取证任务：包含200,000张正常灰度图像，以及用于生成篡改图像的代码。其支持的篡改操作类型包括：加性高斯白噪声（Additive white Gaussian noise，噪声标准差为1）、3×3内核中值滤波（Median filtering）、3×3内核高斯模糊（Gaussian blurring，标准差为0.4）。 针对隐写分析任务：包含160,000张正常灰度图像与隐写图像。其支持的隐写类型包括：最低有效位（Least Significant Bit, LSB）、S-UNIWARD、WOW。