Auto machine learning tools to distinguish between two killer whale ecotypes

In the northwestern Pacific Ocean, there are two ecotypes of killer whales: residents or R-type (fish-eaters) and transients or T-type (mammal-eaters). Most attempts to determine the morphological distinctions between these ecotypes were either based on descriptive variations or utilized approaches that were impractical due to their time-consuming nature or low accuracy. Machine learning algorithms, a subfield of artificial intelligence, show significant potential for image classification. The present study used auto machine learning to differentiate between the dorsal fins of these two killer whale ecotypes using raster images obtained through field surveys. Two auto machine learning platforms were employed: Edge Impulse and Google Cloud AutoML. Both platforms demonstrated high performance. The Edge Impulse platform achieved an accuracy of 90.19%, while the Google Cloud platform achieved an average accuracy of 98.17%. Results show that machine learning stands out as a vital tool for image classification, effectively differentiating ecotypes and confirming that the morphological distinctions between these two ecotypes are not subjective interpretations. Machine learning promises to expand in its uses as an innovative and affordable method for studying the characteristics of cetaceans.