Solo traveler application with snake detection

Abstract

The snakes are not the most famous animals among all other animals. They are perceived as animals should be feared and killed. There are more than 3000 snake species all around the world except in Antarctica, Iceland, Ireland, New Zealand and Greenland. Among these species 600 are venomous and 200 species venomous enough to kill a man. Snakebite is a critical medical emergency that requires quick medical treatment. To make medical treatment, identifying snake is the most vital task. People normally identify snakes based on visual features like body shape, eye shape and color patterns. The knowledge of identification snakes which is not ordinary for many people where only a few experts have this knowledge. This study was focused on creating an automatic snake image classification system that supports mobile and web-based systems. To the best of my knowledge this is the best mobile application that works fully offline and give better accuracy. Convolutional Neural Network was applied to train the snake classification model because Convolutional Neural Network has been obtained great results in image classification. Snake dataset was collected through AICrowd competition. Few Convolutional Neural Network algorithms had been applied on the snake dataset and identified the best algorithm for snake classification. The mobile system was optimized to classify the snakes with and without internet. Therefore, TensorFlow Lite was used to convert the trained image classification model to mobile support format. TensorFlow and Keras was used as the framework for developing and testing the image classification model.