Ornithological Identification from Images Using Convolutional Neural Networks

As society works towards solutions to address environmental concerns, a crucial task emerges: the conservation of species, including avian species, through precise identification and monitoring. This paper shows that it is necessary to develop a quick and efficient way to identify birds in pictures. This project aims to improve the efficiency of bird identification by using a convolutional network to identify 10 distinct species of birds based on their visual representations. The results of the model are good: eventually, the model achieved decent performance indicators, including an accuracy of 0.75, macro average precision of 0.76, and macro average recall of 0.75. These metrics reflect the model’s ability to not only correctly identify birds but also minimize false identifications. Notably, this model demonstrates remarkable performance in classification of common birds such as the American Crow and the California Gull. Moreover, it can also distinguish some nuances between distinct species of birds and distinguish the bird from the background environment.