2018 Looking at People ECCV Satellite Challenge - Track 3 - fingerprint denoising

The following evaluation criteria will be used for this task.

Reconstruction performance: Reconstruction performance will be evaluated on the provided synthetic test set. Participants will be required to reconstruct the degraded fingerprint images in this set by using their developed algorithms and submit the reconstructed fingerprint images. After the submission, the reconstructed fingerprint images will be compared against the corresponding ground-truth fingerprint images in the pixel space to determine the quality of the reconstructions. The results will be quantified as the peak signal-to-noise-ratio (PSNR) and the structural similarity (SSIM).

Matching performance: Matching performance will be evaluated on the provided real test set. Participants will be required to enhance the fingerprint images in this set by using their developed algorithms and submit the enhanced fingerprint images. After the submission, the enhanced fingerprint images will be compared against one another by using a publicly available standard matcher (see baselines) to determine whether they are the impressions of the same fingerprint or not. The results will be quantified as the false acceptance rate (FAR) and the false rejection rate (FRR). 9,730 impostor tests will be conducted to estimate FAR. That is, the nth impression of each of the 140 fingerprints will be matched to the nth impression of the remaining 139 fingerprints (where n will be chosen randomly). 9,240 genuine tests will be conducted to estimate FRR. That is, each of the 12 impressions of the 140 fingerprints will be matched to the remaining 11 impressions of the same fingerprint.