AUTOMATED BIRD SPECIES IDENTIFICATION USING DEEP LEARNING WITH IMAGE AND AUDIO

Abstract

Although watching birds is a popular hobby, identifying their species requires the use of bird books. There are more than 9000 species of birds in the world. Some bird species are rarely discovered, and when they are, prediction is quite challenging.
Additionally, visual recognition of birds by humans is more comprehensible than audible recognition of birds. In order to give birdwatchers a useful tool to appreciate the beauty of birds, we have utilized Convolutional Neural Networks (CNN) to classify bird species as CNNs are a powerful collection of machine learning techniques that have shown to be effective in image processing and sound processing.
This system uses the Caltech-UCSD Birds 200 [CUB-200-2011] and Kaggle dataset for training and evaluating a CNN system for classifying bird species based on image recognition, and several different sound sources for training the sound recognition model.