Fauna Image Classification using Convolutional Neural Network
Abstract
Today, with the increasing volatility, necessity and applications of Artificial Intelligence,
fields like Neural Networks, and its subsets, Machine Learning, and Deep Learning have
gained immense momentum. It has become a data centric model where neural network
developers are “training” the network to be “intelligent” and “independent”. The training
needs softwares and tools such as classifiers, which feed huge amounts of data, analyze
them and extract useful features. These features are then used to observe a pattern and
train the network to use similar data again the next time it is fed data. Convolutional Neural
Network remains to be the most sought-after choice for computer scientists for image
recognition, processing and classification. This paper proposes a fauna image classifier
using convolutional neural network, which will be used to classify images of different
species and animals captured in dense forest environments to achieve desired accuracy,
and aid ecologists and researchers in neural network, artificial intelligence & zoological
domains to further study and/or improve habitat, environmental and extinction patterns. A
convolutional neural network is trained and developed for efficiently classifying these
images with accurate results. Our model was successfully trained with 91.84% accuracy,
and classified images with 99.77% accuracy. Complimentary technologies like VGG16,
TensorFlow, Leaky ReLU, etc. have been used in training the model.