Scaling marine mammals monitoring using AI: A human-in-the-loop solution to analyze aerial datasets

Remote sensing is key for monitoring marine wildlife, offering dynamic solutions for tracking populations like marine mammals. However, it often produces vast datasets that require extensive human effort to analyze. Automatic image analysis using deep learning has been explored, but models face challenges generalizing to new datasets with varying conditions. To address this, a human-in-the-loop approach combines AI with expert biologist input for high-quality analysis. This method was tested on aerial images of belugas in Nunavut, achieving over 90% agreement with expert annotations using only 100 manual inputs. It also saved 96% of the time compared to manual analysis, showing its potential for scaling wildlife monitoring.