CASIA v1.0, CASIA v2.0, Columbia Gray, Columbia Color, NIST 2016(Nimble Challenge 2016 (NC16)), NC17, MFC18, Fantastic Reality, Carvalho, Realistic Tampering, COVERAGE, CoMoFoD

github2024-04-04 更新2024-05-31 收录

下载链接：

https://github.com/greatzh/Image-Forgery-Datasets-List

下载链接

链接失效反馈

官方服务：

资源简介：

本仓库收集了多个用于图像篡改检测和定位的数据集，包括CASIA v1.0、CASIA v2.0等，这些数据集涵盖了不同的篡改类型和处理方式，用于训练和测试图像篡改检测方法。

This repository compiles a variety of datasets for image tampering detection and localization, including CASIA v1.0, CASIA v2.0, among others. These datasets encompass different types of tampering and processing methods, utilized for training and testing image tampering detection techniques.

创建时间：

2023-04-17

原始信息汇总

数据集概述

CASIA v1.0

Forgery Types: Splicing, copy move, removal
Real/Forged Images: 800/921
Image Format: JPEG, TIFF

CASIA v2.0

Forgery Types: Splicing, copy move, removal
Real/Forged Images: 7491/5123
Train/Test Images: 715/100
Image Format: TIFF, JPEG, BMP
Post-processing: Yes
Download: http://forensics.idealtest.org/#/ , https://github.com/namtpham/casia1groundtruth , https://github.com/namtpham/casia2groundtruth
Paper: CASIA Image Tampering Detection Evaluation Database
Year: 2013

Columbia Gray

Forgery Types: Splicing
Real/Forged Images: 933/912
Image Format: BMP
Train/Test Images: 1845, 933 Authentic, 912 Spliced, 128 x 128
Download: https://www.ee.columbia.edu/ln/dvmm/downloads/AuthSplicedDataSet/AuthSplicedDataSet.htm
Paper: A data set of authentic and spliced image blocks
Year: 2004

Columbia Color

Forgery Types: Splicing
Real/Forged Images: 183/180
Train/Test Images: 125/45
Image Format: TIFF
Mask: Yes
Download: Columbia Uncompressed Image Splicing Detection Evaluation Dataset: https://www.ee.columbia.edu/ln/dvmm/downloads/authsplcuncmp/
Paper: Detecting Image Splicing using Geometry Invariants and Camera Characteristics Consistency
Year: 2006

NIST 2016(Nimble Challenge 2016 (NC16)), NC17, MFC18

Forgery Types: Splicing, copy move, removal
Real/Forged Images: 560/564
Train/Test Images: 184/50
Download: NC17: https://www.nist.gov/itl/iad/mig/nimble-challenge-2017-evaluation; MFC18: https://www.nist.gov/itl/iad/mig/media-forensics-challenge-2018; REF: https://mfc.nist.gov/#pills-resources
Paper: MFC Datasets: Large-Scale Benchmark Datasets for Media Forensic Challenge Evaluation
Year: 2019

Fantastic Reality

Forgery Types: Splicing
Real/Forged Images: 16592/19423
Train/Test Images: 12000/1000
Download: http://zefirus.org/MAG
Paper: The Point Where Reality Meets Fantasy: Mixed Adversarial Generators for Image Splice Detection
Year: NeurIPS 19

Carvalho

Forgery Types: splicing
Real/Forged Images: 100/100
Image Format: PNG
Mask: yes
Post-processing: yes
Download: http://www.ic.unicamp.br/tjose/files/database-tifs-small-resolution.zip (BROKEN) http://ic.unicamp.br/~rocha/pub/downloads/2014-tiago-carvalho-thesis/tifs-database.zip
Paper: Exposing digital image forgeries by illumination color classification
Year: TIFS 13

Realistic Tampering

Forgery Types: object insertion , removal
Train/Test Images: 220
Image Format: TIFF
Download: https://pkorus.pl/downloads/dataset-realistic-tampering
Paper: Multi-scale analysis strategies in prnu-based tampering localization
Year: TIFS 16

COVERAGE

Forgery Types: Copy move
Train/Test Images: 100 pairs, 400 x486
Image Format: TIFF
Download: https://github.com/wenbihan/coverage
Paper: COVERAGE — A novel database for copy-move forgery detection
Year: ICIP 16

CoMoFoD

Forgery Types: Copy move
Real/Forged Images: 260/260
Train/Test Images: 260 sets
Post-processing: Yes
Download: https://www.vcl.fer.hr/comofod/
Paper: CoMoFoD — New database for copy-move forgery detection
Year: 2013

MICC F220/F2000

Forgery Types: copy move
Real/Forged Images: 110/110, 440/160, 1300/700
Train/Test Images: 2200
Paper: A SIFT-Based Forensic Method for Copy–Move Attack Detection and Transformation Recovery
Year: TIFS 11

FAU/Manip

Forgery Types: Copy move
Train/Test Images: 48
Post-processing: Yes
Download: http://www5.cs.fau.de/research/data/image-manipulation/index.html
Paper: An Evaluation of Popular Copy-Move Forgery Detection Approaches
Year: TIFS 12

Dresden

Download: http://forensics.inf.tu-dresden.de/ddimgdb/
Paper: The Dresden Image Database for Benchmarking Digital Image Forensics
Year: SAC 10

GRIP

Forgery Types: Copy move
Download: http://www.grip.unina.it/download/prog/CMFD/
Paper: Efficient dense-field copy-move forgery detection
Year: TIFS 15

IMD2020

Forgery Types: Splicing, copy move, removal
Real/Forged Images: 35000/35000
Train/Test Images: 2010
Post-processing: Yes
Download: http://staff.utia.cas.cz/novozada/db/
Paper: An evaluation of popular copy-move forgery detection approaches
Year: WACV 20

In the Wild

Forgery Types: Splicing
Real/Forged Images: -/201
Mask: Yes
Download: https://minyoungg.github.io/selfconsistency/
Paper: Fighting Fake News: Image Splice Detection via Learned Self-Consistency
Year: ECCV 18

DEFACTO

Forgery Types: Splicing, copy move, removal
Real/Forged Images: -/229000
Download: https://defactodataset.github.io/
Paper: http://www.eurecom.fr/en/publication/5973/download/sec-publi-5973.pdf

PS-battles

Forgery Types: Splicing, copy move, removal
Real/Forged Images: 11142/102028
Download: https://github.com/tophatraptor/psdetector.git . Link: https://github.com/dbisUnibas/PS-Battles
Paper: The PS-battles dataset—an image collection for image manipulation detection
Year: CoRR 18

Wild Web

Forgery Types: splicing
Real/Forged Images: 90/9657
Mask: Yes
Image Format: PNG
Paper: Detecting image splicing in the wild (WEB)
Year: ICMEW 15

VIPP Synth

Forgery Types: splicing
Real/Forged Images: 4800/4800
Mask: yes
Image Format: JPEG
Paper: Image Forgery Localization via Block-Grained Analysis of JPEG Artifacts
Year: TIFS 12

VIPP Real

Forgery Types: splicing
Real/Forged Images: 69/69
Mask: manual
Image Format: JPEG
Paper: Image Forgery Localization via Block-Grained Analysis of JPEG Artifacts
Year: TIFS 12

Paper: Image Provenance Analysis at Scale
Year: TIP 18

AbhAS

Forgery Types: splicing
Real/Forged Images: 45/48
Image Format: JPEG
Paper: AbhAS: A Novel Realistic Image Splicing Forensics Dataset
Year: Journal of Applied Security Research 22

MISD

Forgery Types: splicing
Real/Forged Images: 618/300
Mask: Yes
Image Format: JPEG
Post-processing: Yes
Download: https://zenodo.org/records/5525829
Paper: Multiple Image Splicing Dataset (MISD): A Dataset for Multiple Splicing
Year: Data 21

DSO-1

Forgery Types: splicing
Real/Forged Images: 100/100
Mask: Yes
Image Format: JPEG
Download: https://recodbr.wordpress.com/code-n-data/#dso1_dsi1
Paper: Exposing digital image forgeries by illumination color classification
Year: TIFS 13

DIS25k

Forgery Types: splicing
Real/Forged Images: 0/24964
Train/Test Images: 21376/3588
Image Format: JPG
Post-processing: deep image harmonization
Download: Dataset: https://www.kaggle.com/datasets/erentahir/dis25k
Paper: Deep Image Composition Meets Image Forgery
Year: 2024

tampCOCO

Download: https://www.kaggle.com/datasets/qsii24/tampcoco
Paper: CAT-Net
Year: IJCV 22

compRAISE

Download: https://www.kaggle.com/datasets/qsii24/compraise
Paper: CAT-Net
Year: IJCV 22

CocoGlide

Download: https://www.grip.unina.it/download/prog/TruFor/CocoGlide.zip
Paper: TruFor: Leveraging all-round clues for trustworthy image forgery detection and localization
Year: CVPR 23

搜集汇总

数据集介绍

构建方式

该数据集集合了多个图像篡改检测与定位的数据集，涵盖了CASIA v1.0、CASIA v2.0、Columbia Gray、Columbia Color等多个知名数据集。这些数据集通过不同的图像格式（如JPEG、TIFF、BMP等）和处理方式（如是否进行后处理）构建，旨在模拟真实的图像篡改场景，包括拼接、复制移动和移除等篡改类型。每个数据集都经过精心设计，以确保其能够有效支持图像篡改检测算法的研究与评估。

使用方法

使用该数据集集合进行研究时，首先需根据研究目的选择合适的数据集。例如，若研究重点在于图像拼接检测，可以选择Columbia Gray或Columbia Color数据集。随后，研究人员可以根据数据集的格式和处理方式进行预处理，如图像格式转换或后处理操作。最后，利用这些数据集训练和评估图像篡改检测算法，确保算法在不同篡改类型和图像格式下的性能表现。

背景与挑战

背景概述

图像篡改检测与定位是数字图像取证领域的重要研究方向，旨在识别和定位图像中的篡改区域。CASIA v1.0、CASIA v2.0、Columbia Gray、Columbia Color等数据集自2004年起陆续发布，由哥伦比亚大学、CASIA等机构主导，主要研究图像拼接、复制移动和移除等篡改类型。这些数据集的创建为图像篡改检测算法的发展提供了基准，推动了该领域的技术进步。特别是CASIA v2.0，作为早期大规模数据集之一，为后续研究奠定了基础，其影响力延续至今。

当前挑战

图像篡改检测面临多重挑战。首先，篡改技术不断演进，从简单的图像拼接到复杂的深度伪造，检测算法需不断适应新威胁。其次，数据集构建过程中，如何确保样本的真实性和多样性是一大难题。例如，Columbia Gray数据集在构建时未进行后期处理，导致其应用范围受限。此外，不同数据集间的标注标准和格式差异，增加了算法跨数据集迁移的难度。最后，大规模数据集如IMD2020的标注工作耗时且成本高昂，如何高效构建和维护高质量数据集仍是一个开放问题。

常用场景

经典使用场景

在图像篡改检测领域，CASIA v2.0数据集以其丰富的图像样本和多样的篡改类型成为经典。该数据集主要用于训练和评估图像篡改检测算法，特别是针对拼接、复制移动和移除等篡改类型的检测。通过使用CASIA v2.0，研究人员能够开发出更为精确和鲁棒的图像篡改检测模型，从而提升图像取证技术的可靠性。

解决学术问题

CASIA v2.0数据集解决了图像篡改检测中的关键学术问题，即如何有效区分真实图像与经过篡改的图像。通过提供大量经过篡改的图像样本，该数据集帮助研究人员开发和验证新的检测算法，从而推动了图像取证领域的发展。其意义在于，通过提高篡改检测的准确性和鲁棒性，为数字图像的真实性提供了强有力的保障。

实际应用

在实际应用中，CASIA v2.0数据集被广泛用于图像取证、网络安全和媒体真实性验证等领域。例如，在新闻媒体中，该数据集帮助检测和防止虚假图像的传播；在司法取证中，它支持对数字证据的真实性进行鉴定。此外，CASIA v2.0还为社交媒体平台提供了技术手段，以识别和过滤经过篡改的图像内容，维护信息的真实性和可信度。

数据集最近研究