A collection of algorithms to synthetically create scientific images for forensics and integrity analysis.
An in-depth explanation of each algorithm and dataset is described in our research work: Benchmarking Scientific Image Forgery Detectors
The library implements the most type of image tampering functions.
- Image Duplication
- Retouching
- Cleaning
This notebook explains how to apply each type of forgery in a scientific image.
The library also mimics the behavior of images placed in scientific documents, such as compound figures -- with indicative letters and graphs.
There are two possible types of forgeries for compound figures:
-
Intra-panel (forgeries that are isolated within a single panel from the compound figure)
Notebook explaining each type of implemented forgery
-
Inter-panel (forgeries that involve more than one figure panel):
Notebook explaining each type of implemented forgery
Requirements:
To run the notebooks, make sure to install python3.8 and the modules included in the requirements.txt.
Using the implemented library, we created a synthetic dataset dedicated to forensics purposes and scientific integrity.
RSIID Dataset:
Source figures and compound figure templates used to create the tampering dataset:
Both train and test sets have simple and compound figures, organized with the following schematic:
Simple Images
Compound figure
The dataset is distributed under the Attribution 4.0 International (CC BY 4.0) license.
If you use any content from this repository, please cite:
@article{cardenuto_2022,
title={Benchmarking scientific image forgery detectors},
volume={28}, DOI={10.1007/s11948-022-00391-4},
number={4},
journal={Science and Engineering Ethics},
author={Cardenuto, João P. and Rocha, Anderson}, year={2022}
}