CDDB
收藏资源简介:
CDDB(持续深度伪造检测基准)是由苏黎世联邦理工学院和新加坡管理大学等机构共同创建的数据集,专注于研究持续出现的深度伪造检测问题。该数据集收集了来自已知和未知生成模型的深度伪造数据,设计了多种评估方案,以检测模型在面对简单、困难和长期序列的深度伪造任务时的表现。CDDB的应用领域主要集中在提高深度伪造检测的准确性和效率,解决隐私、社会安全和民主等方面的问题。
CDDB (Continuous Deepfake Detection Benchmark) is a dataset co-developed by institutions including ETH Zurich and Singapore Management University, focusing on research into continuous deepfake detection. This dataset collects deepfake data from both known and unknown generative models, and designs multiple evaluation protocols to assess model performance on deepfake detection tasks involving simple, challenging, and long-sequence inputs. The primary application domains of CDDB are centered on improving the accuracy and efficiency of deepfake detection, and addressing issues related to privacy, social security, and democracy.




