Computation of person re-identification using self-learning anomaly detection framework in deep-learning

Abstract

This paper proposes an application of a self-Learning anomaly detection framework in Deep-learning. In this application, both hybrid unsupervised and supervised machine learning schemes are used. Firstly, it takes metadata of the unsupervised data clustering module (DCM). Data clustering module (DCM) analyses the pattern of the monitoring data and enables the self-learning capability that eliminates the requirement of the prior knowledge of the abnormal network behaviors and also has the potential to detect the unforeseen anomalies. Next, we use the self-learning mechanism that transfer pattern learned by the DCM to a supervised data regression and classification module (DRCM) it’s Complexity is mainly related to scalability of supervised learning module. It is more measurable and less time consuming for online anomalies by avoiding excessively usage of the original dataset. It has a density-based clustering algorithm and deep learning, neural network structure-based DCM and DRCM. We are also using an anti-spoofing-based approach for presentation attack detection (PAD). In these approaches, we are mainly detecting a person reidentify and computing without having any false anomalies.