遇见数据集

Reproducibility materials for label-sealed cross-domain reliability auditing of deepfake video detectors

收藏
Zenodo2026-07-20 更新2026-08-01 收录
官方服务:

资源简介:

Sanitized manifests; derived DFDC, DeepfakeTIMIT/VidTIMIT, SDFVD, and HiDF predictions; fixed multiseed and independent-backbone summaries; LSRA/CTSG/ASRG protocols; analysis code; result tables; and deterministic vector-figure sources supporting analysis-level reproduction of the SCI3 manuscript. Raw videos, face crops, feature caches, pretrained weights, trained checkpoints, and complete detector training/preprocessing/inference source are excluded. Licensing: except where a file states otherwise, derived data, documentation, tables, and figures are licensed under CC BY-NC 4.0. Python and PowerShell source-code files are additionally licensed under the MIT License; see LICENSE_CODE_MIT.txt.

提供机构:
Zenodo
创建时间:
2026-07-20
二维码
社区交流群
二维码
科研交流群
商业服务