Open m-langer opened 10 months ago
The dataset consists of four datasets – pls ignore the “PROTECT” one, as these samples are of different type (i.e. dorsal handvein samples instead of fingervein samples as for the other three datasets).
Each folder contains subfolders for the genuine examples (bona fide), the spoofed ones and the synthetic ones:
where GANmethod describes the GAN used to generate the data. Currently, we have cycleGAN for each dataset, but additionally distanceGAN only for the PLUS dataset (so if you should compare generation methods, this is your dataset).
For the synthetic subfolders there are further subfolders where the first one indicates the ID of the variant according to the table on the first page of Description.pdf (see also this file !) Then there are usually 5 subfolders according to the folds used for generation, where the _rs subfolders indicate the images that are resized and the all_rs containing images combined from all those runs.
Each one then again contains a reference subfolder, where the synthetic samples are stored.
An example path would be: PROTECT/spoofed_synthethic_cyclegan/010/3_rs/reference
where the data is from the PROTECT dataset, it is the synthetic spoofs generated by cyclegan, variant 010, fold 3, resized images
DO use all 5 subfolders according to the normal folds, DO NOT use the 5 _rs subfolder DO NOT use the all_rs subfolder
PLUS: 003 and 004 (different generation variants wrt. fold construction, separate experiments, do not mix) SCUT: 007 and 008 (different generation variants wrt. fold construction, separate experiments, do not mix) VERA: 009
In the following, you find a description of how the file names of the samples are composed for better understanding of subject IDs (if you need to separate subjects into training and testing subjects).
The Filename format for the particular datasets and synthetic samples is the following:
All filenames start with a consecutive number separated by a "-" from the original dataset's naming scheme e.g. 001-PLUS-FV3-Laser_PALMAR_001_01_02_01.png where "001" is the consecutive number and "PLUS-FV3-Laser_PALMAR_001_01_02_01.png" is the original filename, according to the format of the datasets:
The filename are encoded using the following structure: [scanner name][DORSAL/PALMAR][session ID][user ID][finger ID]_[image ID].png an example filename is: PLUS-FV3-Laser_PALMAR_001_01_02_01.png
The filenames are encoded using the following structure: c[session ID]_b[sensor_ID]_su[user ID][finger ID]-[image ID].png where
example file name: c1_bF_s_u000_02-000.png
Samples are stored as follow with the following filename convention: full/bf/004-F/004_L2. The fields can be interpreted as
The
Images are labeled as follow: ID_finger_session_shot_light.bmp, where “ID” stands for client's ID, “finger” ranges from 1 to 6 standing for the index, middle and ring finger of right and left hand respectively, “session” stands for session number which can be "0" or "1" “shot” stands for the considered shot number ranging from 0 to 5, and “light” stands for the level of light intensity which can be an integer between 1 and 6.
Images from the same client are regrouped into a single folder labeled.
NOTE: When processing these images, you only need to consider the first two label, i.e. “ID” and “finger”.
All datasets used in the experiments are derived from the databases listed below. The images of those databases are saved in the respective folders. • PLUS • SCUT • VERA • PROTECT
The figure below gives an overview of all used datasets. The datasets are stored in the “Datasets” folder, grouped by their Databases, and named by the ID listed in the figure.
The “Results” folder is grouped by the networks and then by their datasets and experiment number.
An overview of the executed experiments is shown in the figure below.
here are two subfolders for each executed experiment. The “evaluation” and the “output” folder.
First one holds the plots and tables from each single run of the 5-folds and from the all-run. The
“OverviewImage.png” shows four example outputs from the network in comparison to the genuine
and manually spoofed images. The “OverviewTableCompare.tex” holds the important values from
the matching scores from a single run of the 5-folds. “OverviewTableCompareAll.tex” holds the
values from all runs together. The “output” folder holds all images created by the network from all
runs of the 5-folds. There is also a subfolder, containing images combined from all those runs. The
“_rs” folders are containing the resized images from the single runs.
new dataset has no VERA. Which variant(ID) should we use for PROTECT and IDAP
@AIS-22/biometricsystems
Betrifft: LV 911.100 23W 1SSt PS Biometric Systems
Dear Students,
pls find the data @ https://www.cosy.sbg.ac.at/~uhl/Data_prepared.zip
so far, ps only consider original and spoofed data, but NOT the synthetic data, as the directory structure of the latter is not yet fully clarified.
best, AU