Fingerprint representation
This Challenge is available for either contact and contactless fingerprint or both.
Please indicate during registration.
In modern biometric systems, the compactness and the discriminability of feature vectors are fundamental to guarantee high performance in terms of accuracy and speed. Competitors are invited to submit a liveness detection algorithm which returns in addition to the probability of liveness, the feature vector corresponding to the input image with a maximum size of 512 bytes.
The algorithms will be assessed on the basis of system accuracy and system speed on PC with these characteristics:
- Desktop-PC Linux 18.04.1 Ubuntu – Intel® Core™ i9 9900K @ 3.60GHz; RAM 64 GB DDR4 2.933 MHz; Dual NVIDIA® GeForce® RTX 2080 Ti (11GB each)
- Desktop-PC Windows 10 Pro – Intel® Core™ i9 9900K @ 3.60GHz; RAM 64 GB DDR4 2.933 MHz; Dual NVIDIA® GeForce® RTX 2080 Ti (11GB each)
Each submitted algorithm must be a Windows or Linux console application with the following list of parameters:
Synopsis: [nameOfAlgorithm].exe [ndataset] [probeimagesfile][livenessoutputfile] [embeddingsfile]
[nameOfAlgorithm].exe – It is the executable name. Format: Windows or Linux console application (es. .exe)
[ndataset] – It is the identification number of the dataset to analyse. Legend:
- GreenBit
- Dermalog
- Unknown1
- Unknown2
[probeimagesfile] – A text file with the list of absolute paths of each image to analyse. Each image is in the same format of own train set. Example of a [probeimagesfile] called GreenBit_probe.txt:
…
G:\Works\LivDet2025\ProbeImages\GreenBit\0123.png
G:\Works\LivDet2025\ProbeImages\GreenBit\1323.png
G:\Works\LivDet2025\ProbeImages\GreenBit\0033.png
G:\Works\LivDet2025\ProbeImages\GreenBit\0954.png
G:\Works\LivDet2025\ProbeImages\GreenBit\2154.png
…
[livenessoutputfile] -The text file with the liveness output of each processed image, in the same order of “probeimagesfile”. The output is a posterior probability of the live class given the image, or a degree of “liveness” normalized in the range 0 and 100 (100 is the maximum degree of liveness, 0 means that the image is fake). Scores [0, 50) classify fingerprint image as “fake” while scores [50,100] classify fingerprint image as “live”(in Fig.1 in blue “livenessoutput”). In the case that the algorithm has not been able to process the image, the correspondent output must be -1000 (failure to enroll). Example of a [livenessoutputfile] called GreenBit_liveness_results.txt:
…
100
54
-1000
32
78
…
[embeddingsfile] – The .npy or .mat file with the feature vector of each processed image (in the same order of probeimagesfile).
Each parameter related to the dataset configuration must be set before submission. Each competitor can configure his algorithm by the training-set available after the registration. Only Windows or Linux console applications with the above characteristics will be accepted for the competition.
Participants are recommended to send a description of their algorithm.