finger_net is neural network that finds minutiaes on fingerprint image.
Architecture was inspired by MENet. It takes enhaced by gabor filter fingerprint image and returns probability matrix that corresponds to probability whether a given square has minutiae or not.I recommend use this implementation of Gabor filter Model uses Convolutioanl sliding window algorithm to produce the probabality matrix Then minutiaes points are extracted from probabilty matrix.
To compare two fingerprint you need to find corresponding minutiaes. Therefore affine transformation is needed. To do this RANSAC algorithm was used. Determining whether fingerprints are same or not is made by dividing quantity of matching to all points