Bird Species Classification

Problem Statement

Birds prove to be instrumental in checking the health of the earth and indicate climatic change in the environment. In comparison to video based monitoring, audio classification proves to be instrumental because the sounds have the advantage of propagation to long distances without being interfered with objects in between the emitting source and the recording devices. A bird call identification system will be able to monitor species diversity at a specific location and also help us to understand what bird call signals they use in order to communicate with each other for mating purposes or to indicate about the probable danger in the surroundings.

Project Goals

Make a binary classifier that informs whether bird call is present in the audio or not.
Make a multi-classifier that gives information about the species of the bird in the audio.
Combine the first two steps and make a web app that first runs audio through the binary classifier and the audios, that are then predicted as birdcall, are then passed through the multi-classifier to get the species.
The web app will have the functionality to upload audios and get predictions through the models. The annotators can then listen to the audios and change the labels if, according to them the labels are incorrect, which will be used by us to make the models in place even better.