Processed Dataset

[guishengzhanga]

Hi, I am very interested in the 'Deepfake Video Detection Using Generative Convolutional Vision Transformer' research. Could you please provide the processed dataset? Thank you very much!!!

[erprogs]

Hello, thank you for your interest in our deepfake research. To create the dataset, I wrote some custom code to extract frames from the DFDC, FF, Celeb dataset. Here is the CViT faceextract custom code for extracting the DFDC dataset. I used it in the CViT paper.

I used the same code to extract data from FF and Celeb with minor logic change. The only difference is that I used a numbering mechanism to balance the ratio between fake and real videos.

e.g. for FF, I used the first 900 for training, 100 for validation, the rest to testing (something like that). I used similar mechanism for the rest of the dataset.

I do not have permission to redistribute those datasets and their derivatives directly. However, I can provide you with the Json list of the images extracted from the videos.

Tip: there is an extracted face dataset that you can download from the dfdc kaggle website.

I hope I've answered your question.

[guishengzhanga]

Thanks for your response, it's very helpful to me.

[erprogs]

Great!

Here is the link to DFDC images on kaggle: dfdc_150

[guishengzhanga]

Thanks very much! Would you be able to share with me the JSON list of images extracted from the videos?

[erprogs]

Here is the Json file.

deepfake_json.zip