Application of Deep Learning for the Classification of Drowning Incidents

Drowning has consistently garnered extensive attention for public safety. In recent years, with the advancements in deep learning and computer vision, methods using image and video data have been developed to recognize and monitor drowning incidents. Individuals in drowning situations are often incapable of emitting audible distress calls or displaying vigorous struggle. Thus, we used deep learning methodologies to assess definitive drowning incidents. Given the scarcity of publicly available drowning datasets, we investigated drowning behaviors for data collection and simulated drowning scenarios. When image analysis results indicated the likelihood of drowning-related incidents, notifications were dispatched to on-site lifeguards for the final judgment and rescue. By detecting and recognizing drowning behaviors, victims of drowning can be saved.