CNN for Butterfly Classification

In the world of widely available mobile phones and tablets, electronic field guides for the identification of different biodiversity components are emerging as a reality. This project aims to classify butterfly images based on their species using a convolutional neural network in TensorFlow. The goal is to get an accuracy of 80% or above on a butterfly data set. The result of this project includes a testing accuracy of 0.47 by a model that was trained for only 18 hours, i.e. partially trained. Another additional goal is to show that a neural network once trained can be used to identify the species of the butterfly in rather short time, which will make identification and tracking of different butterfly species much easier and efficient.