ARPAN2026/deepfake-detection-dataset-v3

---
language: en
license: mit
task_categories:
- image-classification
pretty_name: Deepfake Detection Dataset V3
size_categories:
- 1K<n<10K
tags:
- computer-vision
- deepfake-detection
- image-forensics
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*
dataset_info:
features:
- name: image
dtype: image
- name: cam_image
dtype: image
- name: cam_overlay
dtype: image
- name: comparison_image
dtype: image
- name: label
dtype:
class_label:
names:
'0': fake
'1': real
- name: confidence_score
dtype: float32
- name: original_caption
dtype: string
- name: cam_caption
dtype: string
- name: query
dtype: string
- name: technical_explanation
dtype: string
- name: non_technical_explanation
dtype: string
splits:
- name: train
num_bytes: 279329873.0531309
num_examples: 474
- name: test
num_bytes: 31333055.946869068
num_examples: 53
download_size: 32345973
dataset_size: 310662929.0
---

# Deepfake Detection Dataset V3

This dataset contains images and detailed explanations for training and evaluating deepfake detection models. It includes original images, manipulated images, confidence scores, and comprehensive technical and non-technical explanations.

## Dataset Structure

The dataset consists of:
- Original images (`image`)
- CAM visualization images (`cam_image`)
- CAM overlay images (`cam_overlay`)
- Comparison images (`comparison_image`)
- Labels (`label`): Binary classification (real/fake)
- Confidence scores (`confidence_score`)
- Image captions (`original_caption`, `cam_caption`)
- Queries and explanations (`query`, `technical_explanation`, `non_technical_explanation`)

## Dataset Statistics
- Total samples: 179 entries
- Train split: 161 samples
- Test split: 18 samples
- Image format: PNG
- Labels: Binary (0 for fake, 1 for real)

## Usage

```python
from datasets import load_dataset

# Load the dataset
dataset = load_dataset("saakshigupta/deepfake-detection-dataset-v3")

# Access the splits
train_data = dataset["train"]
test_data = dataset["test"]

# Example: Get the first sample
sample = train_data[0]
image = sample["image"]
label = sample["label"]
technical_explanation = sample["technical_explanation"]
```

This dataset can be used to:
1. Train deepfake detection models
2. Evaluate detection accuracy
3. Study manipulation patterns
4. Understand detection interpretability
5. Research technical and non-technical aspects of deepfake detection
6. Develop explainable AI systems for image forensics

## Citation

If you use this dataset, please cite:
```
@misc{gupta2024deepfake,
title={Deepfake Detection Dataset V3},
author={Gupta, Saakshi},
year={2024}
}
```