FMIDV: Forged Mobile Identity Document Video dataset
This webpage presents a dataset for copy-move forgeries on the identity documents of MIDV-2020 dataset. The forged samples contain many Similar but Genuine Objects (SGO) which has been shown as a challenge for Copy-Move Forgery Detection (CMFD) algorithms and should be useful in many works in digital forensics research.
Any use of this dataset is required to cite the following reference:
M. Al-Ghadi, Z. Ming, P. Gomez-Krämer, J. -C. Burie, M. Coustaty and N. Sidere, "Guilloche Detection for ID Authentication: A Dataset and Baselines," 2023 IEEE 25th International Workshop on Multimedia Signal Processing (MMSP), Poitiers, France, 2023, pp. 1-6, doi: 10.1109/MMSP59012.2023.10337681.
FMIDV size and access
The dataset has a size of 4,7 GB and is hosted on an FTP server of the University of La Rochelle. Please fill this form for getting access to the dataset. In case of any problem, please contact ([email protected] or [email protected]r).
MIDV-2020 dataset collects identity documents in four categories:
The original sample images obtained from Wikimedia Commons and edited to remove non-persistent data (such as signature, photo, and text field values).
pixels
A photo was taken for each physical document sample, given various conditions and using two smartphones. Half of the photos were captured using Apple iPhone XR, and the other half using Samsung S10.
Each physical document sample was scanned using Canon LiDE 220 and Canon LiDE 300 scanners, in two different modes (1000 for each).
Scanning modes:
For each physical document sample a video clip was captured vertically using Apple iPhone XR and Samsung S10, in a resolution of 2160 × 3840 pixels, with 60 frames per second.
FMIDV dataset consists of 28k forged identity documents for 10 countries based on copy-move forgeries on the identity documents of MIDV-2020 dataset.
For each identity document in the template, photo and scan categories of MIDV- 2020, we have generated 7 forged samples based on copy-move operation.
Copy-move operations were applied on zones of sizes 16 × 16 and 32 × 32 and 64 × 64 pixels; selected randomly.
For 16 × 16 and 32 × 32 pixels copy-move forgeries were applied 2 times for 2
different zones, 2 times for 4 zones, and 2 times for 6 zones.
For 64 × 64 pixels copy-move forgery was applied 1 time for only 2 different zones; because we don't have enough available zones for applying this action for some countries e.g. Finland, Serbia and Slovakia. Moreover, applying copy-move operation on 64 × 64 zones are out of interest as they could be detected by manual inspection (naked eyes).
FMIDV is structed as follows:
The format of any forged identity document in FMIDV is as follows: no._category_Px_Zy.png
no.: presents the sample number; 𝑛𝑜. = [00 − 99]
category: presents the category name that one sample belongs
Px: partition size; 𝑥 = {16, 32, 64}, if 𝑥 = 16 that means that copy-move operation done on zones of 16 × 16 etc.
Zy: presents number of selected zones; 𝑦 = {2, 4, 6}
Dr. Musab Al-Ghadi [email protected]
Dr. Zuheng Ming : [email protected]
Dr. Muhammad Muzzamil Luqman : [email protected]
Dr. Petra Gomez : [email protected]
Pr. Jean-Christophe Burie : [email protected]