Register to Download the Initial Dataset
Building effective AI systems requires learning from high-quality examples. The initial dataset of deepfakes is now available to download.
The Deepfake Detection Challenge invites people around the world to build innovative new technologies that can help detect deepfakes and manipulated media. Identifying manipulated content is a technically demanding and rapidly evolving challenge, so we’re working together to build better detection tools.
October 21
An initial deepfake dataset released
December
The Deepfake Detection Challenge begins
Building effective AI systems requires learning from high-quality examples. The initial dataset of deepfakes is now available to download.
When new forms of misinformation emerge, we need new ways to deal with them. New technologies like deepfakes, where realistic AI-generated videos show real people doing and saying fictional things, are rapidly evolving and detecting them presents a huge technical challenge.
No single organization can solve this on its own. That’s why we’re working together on an ongoing initiative.
It’s important to have data that is freely available for the community to use, with consenting participants and few restrictions on usage. Browse a realistic dataset that includes paid individuals who have agreed to participate in the creation of the dataset for this challenge.
Contribute to this challenge by downloading and ensuring the quality of the dataset and challenge parameters. The full dataset release and the DFDC launch will happen at the Conference on Neural Information Processing system (NeurIPS) this December.