IoT based Human Detection System with Neural Network Classifier

Abstract

Neural network is a programming paradigm which enable a computer to learn from observational data. Neural network was created based on the human neuron. Now a day’s Neural network provide more accurate solution in the field of computer vision such as image recognition, face detection and many other field speech recognition, natural language processing. To make model using neural network, one of the hardest part is collect training data. To train a model a neural network need a huge amount of data. In this study, we introduce a technique how to customize training data so that using same amount of data we can achieve better accuracy. As a neural network model, we have used convolutional neural network. Our problem is human detection. We give an input image in our model our model can predict the input image is human or not. We have collected human face image as positive image and images those not contain human face as negative image. Then we build a convolutional neural network which can predict human or non human. We tune our training data in 2 processes then again train our model which can achieve more accuracy than the previous one. We took 3 type of test datasets and compare those test set result.Finally, we implement our model for IoT based smart door system.