Deepfake-vs-Real-60K


# Deepfake-vs-Real-60K

**Deepfake-vs-Real-60K** is a large-scale image classification dataset designed to distinguish between deepfake and real facial images. The dataset includes approximately **60,000 high-quality images**, comprising **30,000 fake (deepfake)** and **30,000 real** images, to support the development of robust deepfake detection models.

By providing a well-balanced and diverse collection, Deepfake-vs-Real-60K aims to enhance classification accuracy and improve generalization for AI-based deepfake detection systems.

## Label Mappings

- **ID to Label**:
`{0: 'Fake', 1: 'Real'}`
- **Label to ID**:
`{'Fake': 0, 'Real': 1}`

## Dataset Composition

The Deepfake-vs-Real-60K dataset is composed of modular subsets derived from:

- `Deepfakes-QA-Patch1`
- `Deepfakes-QA-Patch2`

These curated subsets ensure high diversity and quality, allowing models trained on this dataset to perform effectively across varied real-world scenarios.

## Key Features

- ~30,000 **Deepfake** images (label `0`)
- ~30,000 **Real** images (label `1`)
- Designed for **image classification tasks**
- Supports **training, evaluation,** and **benchmarking** of deepfake detection models
- Ensures **balanced** class distribution and **high-quality samples**

## Citation

If you use this dataset in your research or project, please cite it as follows:

```bibtex
@misc{prithiv_sakthi_2025,
author = { Prithiv Sakthi },
title = { Deepfake-vs-Real-60K (Revision 1c14d74) },
year = 2025,
url = { https://huggingface.co/datasets/prithivMLmods/Deepfake-vs-Real-60K },
doi = { 10.57967/hf/5313 },
publisher = { Hugging Face }
}
```

## License

This dataset is licensed under the **Apache License 2.0**.

For more details, see the [license](https://www.apache.org/licenses/LICENSE-2.0).

## Dataset Page

Explore and download the dataset here:
[https://huggingface.co/datasets/prithivMLmods/Deepfake-vs-Real-60K](https://huggingface.co/datasets/prithivMLmods/Deepfake-vs-Real-60K)


# Deepfake-vs-Real-60K（深度伪造vs真实图像60K数据集）
**Deepfake-vs-Real-60K** 是一款大规模图像分类数据集，旨在区分深度伪造（deepfake）人脸图像与真实人脸图像。本数据集包含约**60000张高质量图像**，其中**30000张为深度伪造样本**，**30000张为真实样本**，旨在支撑鲁棒性深度伪造检测模型的研发。

通过提供均衡且多样化的样本集合，本数据集旨在提升基于AI的深度伪造检测系统的分类精度与泛化能力。

## 标签映射
- **ID到标签映射**：`{0: 'Fake', 1: 'Real'}`
- **标签到ID映射**：`{'Fake': 0, 'Real': 1}`

## 数据集构成
本数据集由以下模块化子集组合而成：
- `Deepfakes-QA-Patch1`
- `Deepfakes-QA-Patch2`
这些经过精心筛选的子集保证了样本的高多样性与高质量，使得基于该数据集训练的模型能够在多样的真实场景中实现高效表现。

## 核心特性
- 约30000张**深度伪造**图像（标签为`0`）
- 约30000张**真实**图像（标签为`1`）
- 专为**图像分类任务**设计
- 支持深度伪造检测模型的**训练、评估与基准测试**
- 保证**类别分布均衡**与**样本质量上乘**

## 引用格式
若您在研究或项目中使用本数据集，请按以下格式引用：
bibtex
@misc{prithiv_sakthi_2025,
author = { Prithiv Sakthi },
title = { Deepfake-vs-Real-60K (Revision 1c14d74) },
year = 2025,
url = { https://huggingface.co/datasets/prithivMLmods/Deepfake-vs-Real-60K },
doi = { 10.57967/hf/5313 },
publisher = { Hugging Face }
}

## 授权协议
本数据集采用**Apache许可证2.0（Apache License 2.0）**授权。
更多详情请参阅[授权协议](https://www.apache.org/licenses/LICENSE-2.0)。

## 数据集页面
您可在此处浏览并下载本数据集：
[https://huggingface.co/datasets/prithivMLmods/Deepfake-vs-Real-60K](https://huggingface.co/datasets/prithivMLmods/Deepfake-vs-Real-60K)