Participants can submit any pretrained model trained on their choice of fingerprint data.
The participants will provide Python ready-to-run scripts of their relevant algorithms. The submitted algorithm will be re-run to ensure reproducibility and a fair comparison.
Open source fingerprint recognition solutions are not allowed to be submitted, unless it is developed by the same participant team.
The Python scripts should access a folder of latent fingerprint images for comparison, the structure of the input folder is exactly the same as the LFIW Dataset used for development and training, only the subject number and subject name are different.
The Python scripts should output a comparison scores csv file formulated as:
image1_path, image2_path, comparison_score
Each team can submit up to 2 different solutions. If two solutions are submitted (per-team), the solution that achieved the best performance will be considered.