daisy-zhang / awesome-deepfakes-detection Goto Github PK

Thanks for your work. May I ask why the comparison content about the benchmark of all the SOTA methods was cancelled in the update after September?

Thank you for this repository, it really helped me a lot!!! ^^

paper: https://www.sciencedirect.com/science/article/pii/S0167865524000990
code: https://github.com/gleporoni/rgbd-depthfake

ECCV 2022 UIA-ViT: Unsupervised Inconsistency-Aware Method based on Vision Transformer for Face Forgery Detection 的代码链接如下：https://github.com/wany0824/UIA-ViT。希望可以为这个仓库尽一点绵薄之力

Thanks for the great work, how is the cross domain for forgery detection done? I found many papers trained on ff++ and tested on cdf and dfdc, how is this achieved? Is it 720 videos selected from each of the four subsets of ff++? I know how the separate subsets are trained, but I don't know how to aggregate them, looking forward to your reply, thanks.

你好，关于FF++的数据集预处理我有些疑惑，我注意到官方提供了划分json 文件，里面是些数字。但是造假视频一般有两个数字表示源视频或者目标视频，那到底是按照源视频的编号他是否属于训练集还是有别的什么方法，以及是否有讨论群之类的啊？

Title : Learning on Gradients: Generalized Artifacts Representation for GAN-Generated Images Detection
Pdf : https://openaccess.thecvf.com/content/CVPR2023/papers/Tan_Learning_on_Gradients_Generalized_Artifacts_Representation_for_GAN-Generated_Images_Detection_CVPR_2023_paper.pdf
Code: https://github.com/chuangchuangtan/LGrad

Thank you very much. Do you know that there are some papers on fake face detection with graph convolutional neural network？

I do not find paper called Detecting and Grounding Multi-Modal Media Manipulation and Hierarchical Fine-Grained Image Forgery Detection and Localization in https://openaccess.thecvf.com/CVPR2023. Could you check it again?

Could you please list some of the papers and open source projects that currently have the best results in deepfake detection in this project? This will be of great help to my study.

大佬你好，我想请问一下如何将这些数据集划分成训练集与测试集。将一段视频的前80%作为训练集，后20%作为测试集这种划分方式可行吗？（感觉不太好，这样就好像把训练集的数据拿到测试集中了）

Hello Daisy, Awesome work on this repo. may i request one thing can you add formal definition of the sections(generalization, Interpretability etc). So it's easier to understand the sections more.

Thanks for such a comprehensive list.

I noticed "Artificial Fingerprinting for Generative Models: Rooting Deepfake Attribution in Training Data", ICCV 2021" was mentioned twice in the ICCV paper list. I think one can be removed.

Thanks.

很抱歉再次打扰您，我最近尝试下载FF++ raw格式的数据集，但是遇到了如下图的问题。我更换了不同计算机也调整了一些设置但是都没有效果。请问还有别的方法可以下载FF++ raw格式的数据集吗

Can you provide the code?