Classifying Endangered Species in High-Risk Areas Using Deep Learning

Endangered animals are protected by national and international regulations as they are part of the environmental, cultural and genetic heritage. Some of these species are difficult to identify and monitor in the wild, hence very little information and data are available about them. Whenever an organization's actions impact this type of species, it can receive huge fines. Despite this situation, there are currently no specific automated methods to accurately identify this type of animal, using small image datasets. This research introduces the use of Deep Learning techniques to address a real environmental problem related to the classification of endangered wildlife that lives within the area of influence of large mining projects. Small datasets were used because there are no public databases available for the target species. The overall model achieved high accuracy in classifying images of different quality and those containing high levels of noise, reaching an average accuracy and F1-score greater than 0.97.