On the Use of YOLO-NAS and YOLOv8 for the Detection of Sea Lions in the Galapagos Islands

Sea lions (Zalophus Wollebaeki) are a protected species, and effective monitoring is crucial for habitat preservation and behavioral studies. However, manual sea lion counting is laborious and error-prone. In this paper, we explore the use of two standard convolutional neural network models (YOLO-NAS and YOLOv8) for sea lion detection as a preliminary step towards automating the counting process. For this purpose, a data set of images and videos of sea lions was collected in their natural environment in the Galapagos Islands. The results demonstrate that both models exhibit promising detection capabilities, successfully identifying almost all sea lions in the images. In particular, YOLOv8 shows to be more reliable in the detection of sea lions under challenging and complex conditions, while YOLO-NAS excels in the identification of a larger number of individuals, including those of a smaller size. These findings pave the way for future developments in automated sea lion counting tools, streamlining conservation efforts, and advancing our understanding of this protected species.