This article delves into the architecture of a system tailored for recognizing individuals at risk of drowning at sea, leveraging advanced computer vision techniques to ensure beachfront safety. The system capitalizes on a camera mounted on a lifeguard tower, capturing a comprehensive view of the beach vicinity. The proposed method integrates two stages: 1) Sea segmentation via the deep learning architecture, SegNet, trained to identify marine portions occupied by swimmers, further enhanced by the VGG16 model to extract intricate image features, and 2) Precise individual detection within the sea using the YOLO framework and the Vision Transformer model, trained to scrutinize sequential frames and detect anomalies suggestive of drowning. This approach underwent rigorous evaluation, signaling potential as a transformative measure in coastal safety protocols, aiming to substantially mitigate drowning-associated hazard