Deep Learning Framework for Detection of an Illicit Drug Abuser Using Facial Image

Abstract

The detection of an illicit drug abuser by analyzing the subject’s facial image has been an active topic in the field of machine learning research. The big question here is up to what extent and with what accuracy can a computer model help us to identify if a person is an illicit drug abuser only by analyzing the subject’s facial image. The main objective of this paper is to propose a framework which can identify an illicit drug abuser just by giving an image of the subject as an input. The paper proposes a framework which relies on Deep Convolutional Neural Network (DCNN) in combination with Support Vector Machine (SVM) classifier for detecting an illicit drug abuser’s face. We have created dataset consisting of 221 illicit drug abusers’ facial images which present various expressions, aging effects, and orientations. We have taken random 221 non-abusers’ facial images from available dataset named as, Labeled Faces in the Wild (LFW). The experiments are performed using both datasets to attain the objective. The proposed model can predict if the person in an image is an illicit drug abuser or not with an accuracy of 98.5%. The final results show the importance of the proposed model by comparing the accuracies obtained in the experiments performed.