arXiv [] 27 Aug 2020
1 We propose a novel CNN architecture to extract fea ture from fingerprint images by combining texture minutiae and frequencyspectrum 2 We propose a minutia attention module to pay atten tion to the position of the minutiae to extract global and local texture features 3 We propose a new data augmentation method specific to
GitHub jakubarendac/fingerflow FingerFlow is
Skeleton minutiae a part of algorithm transparency are minutiae obtained directly from ridge and valley skeletons during feature extraction specifically from the last filtering stage of both skeletons Only skeleton minutiae with exactly one connected ridge or valley are considered Skeleton minutiae from ridge skeleton become endings