ForgeryNet
收藏资源简介:
ForgeryNet是一个大规模的人脸伪造数据集,由商汤科技研究创建,包含数千个主题、多种操纵方法和多样的重新渲染过程。数据集大小为290万张图像和221,247个视频,数据来源于四个不同的数据集。创建过程中,采用了15种图像级操纵方法和8种视频级操纵方法。ForgeryNet的应用领域包括图像和视频分类、空间和时间伪造定位,旨在解决真实世界中的数字伪造问题。
ForgeryNet is a large-scale face forgery dataset developed by SenseTime Research. It encompasses thousands of subjects, a wide range of manipulation techniques, and diverse re-rendering procedures. The dataset comprises 2.9 million images and 221,247 videos sourced from four distinct datasets. During the construction of this dataset, 15 image-level manipulation methods and 8 video-level manipulation methods were adopted. Its application domains include image and video classification, spatial and temporal forgery localization, aiming to address digital forgery issues in real-world scenarios.

- 1ForgeryNet: A Versatile Benchmark for Comprehensive Forgery Analysis商汤科技研究 · 2021年



