Customized deep neural network model for autonomous and efficient surveillance of wildlife in national parks

Abstract

As National parks are one of the best places of attraction to the tourists, the so-called abode of the animals has to be protected due to the frequent intervention of human beings. On the other hand, the lives of human beings have also to be insulated from the sudden and vulnerable attack of the animals. When it comes to ‘life’, it becomes a vital factor for humans and animals. Therefore, here arises a dire need to create a system which can protect the life and safety of human beings as well as animals. However, when it comes to mutual safety of both, there has not been an effective solution till date that can practically be implemented in a national park or a sanctuary. In this scenario, ‘Video Object Identifying’ using Machine Learning Technique has been proven to render promising results in terms of accuracy and performance. Classifying and identifying objects using Customized Convolutional Neural-Network Model in order to detect the multi-class objects (vehicle and animal) become the aim of this present research. In this research, five categories of data sets namely lion, elephant, tiger, car and motorcycle have been used for training and testing. The custom model for video object detection renders with an accuracy of 82% which proves its efficiency in detecting the multi class datasets.