Celeb-DF++：面向可泛化取证的大规模、具挑战性视频 DeepFake 基准
Celeb-DF++: A Large-scale Challenging Video DeepFake Benchmark for Generalizable Forensics
AI总结：
本文提出大规模视频 DeepFake 基准 Celeb-DF++，覆盖三类伪造场景与 22 种生成方法，并通过协议评估 24 种检测器以衡量可泛化取证能力。
摘要：
AI 技术的快速发展显著增加了网络上流传的 DeepFake 视频的多样性，给“可泛化取证”（generalizable forensics）带来了紧迫挑战，即用单一模型检测多种此前未见的 DeepFake 类型。应对这一挑战需要不仅规模庞大、而且伪造类型丰富的数据集。然而，大多数现有数据集尽管具备规模，却只包含有限种类的伪造类型，不足以开发可泛化的检测方法。因此，我们在早期 Celeb-DF 数据集的基础上，提出 Celeb-DF++，这是一个全新的大规模、具挑战性的视频 DeepFake 基准，专门面向可泛化取证挑战。Celeb-DF++ 涵盖三种常见伪造场景：换脸（Face-swap，FS）、面部重演（Face-reenactment，FR）和说话人脸（Talking-face，TF）。每种场景都包含大量高质量伪造视频，这些视频共由 22 种不同的近期 DeepFake 方法生成。这些方法在架构、生成流程和目标面部区域方面各不相同，覆盖了现实环境中最常见的 DeepFake 情形。我们还提出了用于评估 24 种近期检测方法泛化能力的评测协议，揭示了现有检测方法的局限性以及新数据集的难度。
英文摘要：
The rapid advancement of AI technologies has significantly increased the diversity of DeepFake videos circulating online, posing a pressing challenge for \textit{generalizable forensics}, \ie, detecting a wide range of unseen DeepFake types using a single model. Addressing this challenge requires datasets that are not only large-scale but also rich in forgery diversity. However, most existing datasets, despite their scale, include only a limited variety of forgery types, making them insufficient for developing generalizable detection methods. Therefore, we build upon our earlier Celeb-DF dataset and introduce {Celeb-DF++}, a new large-scale and challenging video DeepFake benchmark dedicated to the generalizable forensics challenge. Celeb-DF++ covers three commonly encountered forgery scenarios: Face-swap (FS), Face-reenactment (FR), and Talking-face (TF). Each scenario contains a substantial number of high-quality forged videos, generated using a total of 22 various recent DeepFake methods. These methods differ in terms of architectures, generation pipelines, and targeted facial regions, covering the most prevalent DeepFake cases witnessed in the wild. We also introduce evaluation protocols for measuring the generalizability of 24 recent detection methods, highlighting the limitations of existing detection methods and the difficulty of our new dataset.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
