A novel hybrid deep belief Google network framework for brain tumor classification

Abstract

Within the fields of law enforcement and forensics applications, latent fingerprints have garnered a lot of interest from researchers. The need from the general public for these uses may be what propels biometrics research forward. Although a lot of work has gone into building techniques for latent fingerprint classification, there are still many difficult issues to solve low quality pictures, segmentation, noise, and intra class variations in that field. To overcome the above difficulties, proposed an Automated Latent Fingerprint Recognition framework in this research using strategies for latent fingerprint pre-processing, feature extraction, and matching. A candidate fingerprint's salient minutiae, which give each fingerprint its individuality and distinguish it from others, are first identified and described, followed by their relative placement in the candidate fingerprint and previously saved fingerprint templates. The experimental analyses using publicly accessible low-quality Latent partial fingerprints was taken from MSU PrintsGAN datasets show that the proposed framework achieves an average equal error rate (EER) value of 0.254 and TAR@FAR achieves 91.43 which is contrasted to various existing approaches.