DDL
收藏资源简介:
DDL数据集是一个包含超过180万伪造样本的大规模深度伪造检测和定位数据集,涵盖了75种不同的深度伪造方法。该数据集的设计包括四个关键创新点:多样化的伪造场景、全面的深度伪造方法、不同的操作模式以及精细的伪造注释。DDL数据集不仅为复杂的真实世界伪造提供了一个更具挑战性的基准,还为构建下一代深度伪造检测、定位和可解释性方法提供了重要支持。
The DDL Dataset is a large-scale deepfake detection and localization dataset containing over 1.8 million forged samples, covering 75 distinct deepfake generation methods. The design of this dataset features four key innovations: diverse forgery scenarios, comprehensive coverage of deepfake techniques, varied operating modes, and precise forgery annotations. The DDL Dataset not only provides a more challenging benchmark for complex real-world forgeries but also offers critical support for developing next-generation deepfake detection, localization, and interpretability methods.




