DDL: A Large-Scale Dataset for Deepfake Detection and Localization in Diversified Real-World Scenarios
The DDL dataset was created to improve deepfake detection and localization by addressing limitations in existing benchmarks. It provides over 1.4 million forged samples covering 80 different deepfake methods across various visual and audio modalities, complete with fine-grained spatial and temporal annotations, allowing researchers to build more robust detection tools.