ID测试集

ID测试集是由西安交通大学等机构创建的一个高质量、多样化的数据集，用于深度伪造检测的评估。该数据集包含976个伪造视频和2348个真实视频，涵盖了多种伪造方法和常见的干扰类型，旨在模拟真实世界中的媒体环境。数据集的创建过程包括从公开数据集和私有数据集中选择难以被人类和检测算法区分的样本，确保了数据集的挑战性和实用性。ID测试集主要用于评估深度伪造检测算法的鲁棒性和实用性，以解决当前深度伪造技术带来的隐私和安全问题。

The ID Test Set is a high-quality, diverse dataset developed by institutions including Xi'an Jiaotong University for deepfake detection evaluation. It comprises 976 forged videos and 2348 authentic videos, covering a wide range of forgery methods and common interference types, with the goal of simulating real-world media environments. The dataset was constructed by selecting samples that are difficult for both human evaluators and detection algorithms to distinguish from both public and private datasets, ensuring its challenging nature and practical utility. The ID Test Set is primarily used to assess the robustness and practicality of deepfake detection algorithms, addressing the privacy and security issues arising from current deepfake technologies.