AI-vs-Deepfake-vs-Real

# **AI vs Deepfake vs Real**

**AI vs Deepfake vs Real** is a dataset designed for image classification, distinguishing between artificial, deepfake, and real images. This dataset includes a diverse collection of high-quality images to enhance classification accuracy and improve the model’s overall efficiency. By providing a well-balanced dataset, it aims to support the development of more robust AI-generated and deepfake detection models.

# **Label Mappings**
- **Mapping of IDs to Labels:** `{0: 'Artificial', 1: 'Deepfake', 2: 'Real'}`
- **Mapping of Labels to IDs:** `{'Artificial': 0, 'Deepfake': 1, 'Real': 2}`

This dataset serves as a valuable resource for training, evaluating, and benchmarking AI models in the field of deepfake and AI-generated image detection.

# **Dataset Composition**

The **AI vs Deepfake vs Real** dataset is composed of modular subsets derived from the following datasets:

- [open-image-preferences-v1](https://huggingface.co/datasets/data-is-better-together/open-image-preferences-v1)
- [Deepfakes-QA-Patch1](https://huggingface.co/datasets/prithivMLmods/Deepfakes-QA-Patch1)
- [Deepfakes-QA-Patch2](https://huggingface.co/datasets/prithivMLmods/Deepfakes-QA-Patch2)

The dataset is evenly distributed across three categories:
- **Artificial** (33.3%)
- **Deepfake** (33.3%)
- **Real** (33.3%)

With a total of **9,999 entries**, this balanced distribution ensures better generalization and improved robustness in distinguishing between AI-generated, deepfake, and real images.

# **AI生成图像 vs 深度伪造（Deepfake）图像 vs 真实图像**

**AI生成图像 vs 深度伪造（Deepfake）图像 vs 真实图像**是一款面向图像分类任务的数据集，用于区分AI生成图像、深度伪造图像与真实图像。本数据集收录了多样化的高质量图像，旨在提升分类任务的准确率，并优化模型的整体运行效率。通过构建均衡分布的数据集，本数据集旨在助力更鲁棒的AI生成内容与深度伪造图像检测模型的研发。

# **标签映射规则**
- **标签ID与类别的映射关系**：`{0: 'AI生成图像', 1: '深度伪造图像', 2: '真实图像'}`
- **类别与标签ID的映射关系**：`{'AI生成图像': 0, '深度伪造图像': 1, '真实图像': 2}`

本数据集可作为深度伪造与AI生成图像检测领域中，用于AI模型训练、评估与基准测试的宝贵资源。

# **数据集构成**

**AI生成图像 vs 深度伪造（Deepfake）图像 vs 真实图像**数据集由以下公开数据集衍生的模块化子集构成：

- [open-image-preferences-v1](https://huggingface.co/datasets/data-is-better-together/open-image-preferences-v1)
- [Deepfakes-QA-Patch1](https://huggingface.co/datasets/prithivMLmods/Deepfakes-QA-Patch1)
- [Deepfakes-QA-Patch2](https://huggingface.co/datasets/prithivMLmods/Deepfakes-QA-Patch2)

本数据集在三个类别中实现了均匀分布：
- **AI生成图像**（占比33.3%）
- **深度伪造图像**（占比33.3%）
- **真实图像**（占比33.3%）

本数据集总样本量为**9999条**，这种均衡的分布设计能够提升模型在区分AI生成图像、深度伪造图像与真实图像时的泛化能力与鲁棒性。

搜集汇总

背景与挑战

背景概述

AI-vs-Deepfake-vs-Real 是一个用于图像分类的数据集，专门区分人工生成、深度伪造和真实图像。它包含9,999个条目，三类图像均衡分布各占33.3%，基于多个子数据集构建，旨在提升AI生成和深度伪造检测模型的准确性和鲁棒性，为相关研究提供训练和评估资源。