Proc. 1 st International Conference on Machine Learning and Data Engineering (iCMLDE2017) Page 108 Proc. 1 st International Conference on Machine Learning and Data Engineering (iCMLDE2017) 20-22 Nov 2017, Sydney, Australia ISBN: 978-0-6480147-3-7 Automatic Ear Detection using Deep Learning MD Moniruzzaman 1 and Syed Mohammed Shamsul Islam 2 1 Postgraduate Student, School of Science, Edith Cowan University, Joondalup, WA 6027, Australia 2 Lecturer, School of Science, Edith Cowan University, Joondalup, WA 6027, Australia Corresponding author’s E-mail: syed.islam@ecu.edu.au Abstract Ear based biometric identification can be the solution for instance such as surveillance where other biometric traits are simply very hard to access. Although many semi-automatic approaches have been made to detect ear and use it for human recognition purpose, most of them are based on feature extraction or shallow machine learning approaches. Very few approaches those used deep neural network architectures are either having less hidden layers, or a combination of deep neural network and feature extraction classifiers or already trained complex deep convolutional network. In this research approach, a deep but simple raw convolutional neural network have been used to detect ear from an ear and non-ear environment which is the initial part of ear based biometric implication. Data-Augmentation have also used and a comparative analysis have been done for original data set and Augmented dataset. Using this deep but from scratch architecture trained by 792 original images we achieved promising output which show larger data could achieve higher accuracy. Keywords: Ear detection, Deep Learning, Convolutional Network, Data augmentation. 1. INTRODUCTION During the course of last decade, researchers are focusing actively on biometrics research due to the growing need for automatic authentication of human individuals. In biometric identification, ear based authentication and human recognition proved itself a novel research field as suggested by Zhang and Mu (2017). The potential, credentials and possibilities of using human ear for recognition and human identification was proposed by Bertillon (1890). According to Jain et al (2007) and Omara et al (2017), to satisfy the traditional personal authentication properties, a biometric trait should be unique, universal, permanent and easily collectible. The uniqueness of the helix part of human ear has been proven by Iannarelli (1989). Not only helix, anti-helix, which is a parallel part of helix with a distinctive hairpin-bend shape also proved prominent by Hurley et al (2007). Human facial expression doesn’t have any effect on ear, with age the change is negligible and the background is always predictable as ear is fixed firmly on the side of human head. Collection of ear image does not create a sense of anxiety and no hygiene issue raises as the collection process is touch-less. Moreover, as the size of the ear is significantly larger than other biometric traits like finger print, iris, retina, it proves more accessible while the incident is associated with criminal act. Present Artificial Neural Network based ear detection and recognition approaches can be broadly categorized in to 3D point cloud based approaches or 2D image data approaches. This research work deals with only 2D images. One of the earliest approaches of using Neural Network based Ear recognition was experimented by Galdámez et al (2014). In their approach they used Speeded Up Robust Features (SURF) and Fisher Linear Discriminant Analysis (LDA) as the input of their two neural networks. On their later approach Galdámez et al. (2016) used deep Convolutional neural networks (CNN) based ear recognition system to identify a person by the ear image. Tian and Mu (2016) designed a neural network having three