Each submission should be limited to one Python script called submission_xyz_nnn.py where
xyz is the participant identifier, and nnn is the algorithm identifier (if the participant submits more than one algorithm). The script should accept two command line input arguments: a path to the input CSV file, and a path to the output CSV file.
With the submission Authors might provide:
requirements.txtfile with list of libraries required to run the submission- additional artefacts (ex. model file) and config files for the algorithm, handling of the artifacts should be fully encapsulated in the submission python script
Example command line execution:
python submission_ND_001.py input.csv output.csv
All test images will be coded as lossless, 8 bit 640x480 pixel grayscale PNG, i.e., every pixel is represented by a single sample that is a grayscale level, where 0 is black and 255 is white. Further information on PNG format can be found in RFC2083. Images of authentic irises will conform to the IMAGE_TYPE_VGA image type, as defined in ISO/IEC 19794-6:2011, yet analogous conformance for spoof iris images is not guaranteed.
After registration the participants will obtain an example image representing each task in the competition to make sure that the image format is known.
Input CSV file contains columns with index and paths to an ISO compliant image (see sample.csv for an example CSV). Output CSV file should contain columns with index, file path and the liveness score.
The score should represent the probability (floating point number) between 0.0 and 1.0 that the
image is a presentation attack, where 0.0 means certainly not an attack, and 1.0 means certainly
an attack. If the image cannot be processed, return None.
Before submitting your algorithm validate your submission with submission_validation.py script.
Example command line execution:
python submission_validation.py submission_ND_001.py
The script will print VALID SUBMISSION message for submission that meets the requirements. If the submission is incorrect, it will output the specific errors.
We will use GPU machines to evaluate submissions as specified below. The machines will have standard software packages available (such as Pytorch, Keras, OpenCV, etc.) and we will make an effort to install specific versions of these packages as specified by the participants.
Machine used in Task 1 evaluations:
CPU: Intel Core i7-10700K 8-Core 3.80 GHz
RAM: 64 GB DDR4 3200 MHz
GPU: NVIDIA RTX 4090 24GB
System: Windows 10 Pro
Machine used in Task 2 evaluations:
CPU: AMD Ryzen Threadripper 3960X - 24-Core - 3.8 GHz / 4.5GHz Boost
RAM: 256GB DDR4 3200MHz (32GB x 8)
GPU: NVIDIA RTX 3090 24GB
System: Ubuntu 22.04
Machines used in Task 3 evaluations:
CPU: Intel(R) Core(TM) i7-10700 CPU @ 2.90GHz
RAM: 48 GB
System: Windows 10 Enterprise
CPU: AMD Ryzen 7 5800X 8-Core
RAM: 64 GB
GPU: NVIDIA GeForce RTX 3090 (24,576 MiB)
System: Ubuntu 22.04.4 LTS