IMD2020: A Large-Scale Annotated Dataset Tailored for Detecting Manipulated Images
Description & Download
-
IMD2020 Real-Life Manipulated Images
This dataset contains 2,010 real-life manipulated images downloaded from the Internet. Corresponding real versions of these images are also provided. Moreover, there is a manually created binary mask localizing the manipulated area of each manipulated image.
You can download this whole dataset HERE.
-
IMD2020 Large-Scale Set of Real Images
This dataset contains 35,000 real images captured by 2,322 camera models. All of the images were manually reviewed and those having obvious traces of digital manipulation discarded.
You can download this dataset here:
-
IMD2020 Large-Scale Set of Inpainting Images
You can download this dataset here:
-
IMD2020 Guaranteed Set of Images (32 unique cameras)
This dataset contains 2,759 real images formed by 32 unique cameras (19 camera models) designed for sensor noise analysis (PRNU, etc.).
You can download this dataset here:
Citation
If you use this databases, please cite the following paper:
@INPROCEEDINGS{Novozamsky_2020_WACV,
author = {Novozamsky, Adam and Mahdian, Babak and Saic, Stanislav},
title = {IMD2020: A Large-Scale Annotated Dataset Tailored for Detecting Manipulated Images},
booktitle = {2020 IEEE Winter Applications of Computer Vision Workshops (WACVW)},
year = {2020},
month = {March},
pages = {71-80}
}
Contact
Please contact Adam Novozamsky (novozamsky (at) utia.cas.cz) for questions about the dataset.
Acknowledgements
This project has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement No 825227.
Designed and managed by ZOI UTIA