A Novel Deep Learning Model for Recognition of Endangered Water-Bird Species

Abstract

Given its location on the migration route of the Western Palearctic, the complex of wetlands of El-Kala (North-East Algeria) forms the most important and diverse area of the Mediterranean for migratory birds in the Maghreb. The knowledge of these birds allows one to acquire crucial information on the state of health of considered environments as well as annual statistics of this population. Some of which are threatened with extinction. Because of the dense vegetation, the main feature characterizing the birds' habitat, the identification of bird species from their images is made a complicated task. In addition, there is a high degree of similarity between classes and features. In this paper and in order to solve these problems, a new method named DarkBirdNet based on deep learning has been developed. This method is derived from the predefined DarkNet53 model and aims at detecting and classifying bird species in Algeria.