Real-time detection of marine organisms is essential for effective ocean surveillance, ecological research, and marine conservation. This study proposes a YOLO-based object detection approach for the fast and accurate identification of marine life in underwater environments. By utilizing the single-stage architecture of YOLO, the system achieves high detection speed without compromising accuracy. The model is trained to recognize various marine species, including fish, turtles, and other underwater organisms. To overcome underwater imaging challenges such as poor illumination, color attenuation, and background noise, appropriate preprocessing techniques are applied to enhance input data quality. The trained model processes live video streams and performs real-time inference with low computational latency. Experimental evaluation shows that the proposed method delivers reliable detection performance and real-time efficiency, making it suitable for deployment on embedded systems, underwater robots, and autonomous marine vehicles. The proposed framework contributes to continuous marine ecosystem monitoring and supports data-driven conservation strategies.
K et al. (Thu,) studied this question.