«Shazam for birds» has been created in St. Petersburg State University
Oleg Kenunen created a mobile application based on a neural network with two main purposes. The first was to help amateur ornithologists and the second — to somewhat automate the work of professional ornithologists.
The application recognises just twenty species of birds so far (including the nightingale, finch, and thrush). But it is already capable of correctly distinguishing one species from another with more than 90 percent accuracy, even before the field trials have begun. This is not so bad, given that the recordings are made with a smartphone, and microphones in those are not designed to capture sources of sound which are far from the device. The developer of the application notes that the accuracy can be greatly increased if the voice library would be continually updated, thus continuing to train the neural network. The more the application analyses records, the higher the recognition accuracy.