Skip to content

Dataset · nuriachandra

2024 real deepfake multimodal benchmark

nuriachandra/Deepfake-Eval-2024

Manually annotated video, audio, and images for evaluating deepfake detectors in real-world conditions.

What's inside

Contains 44 hours of video, 56.5 hours of audio, and 1,975 manually annotated images for real and fake content. Materials were gathered from 88 different web sources across 52 languages and cover modern generation technologies. The total volume ranges between 1,000 and 10,000 examples.

What you can do with it

  • Evaluating and testing multimodal models for deepfake detection
  • Benchmarking fake audio and video classifiers on real web data
Size
1K–10K rows
Downloads this month
2.7K
Likes
69
Access
on request on Hugging Face

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("nuriachandra/Deepfake-Eval-2024")