UOD-YOLO: a lightweight real-time model for detecting marine organisms

Introduction Underwater organism detection is critical for intelligent aquaculture and marine ecological monitoring, but real-time deployment on resource-constrained platforms remains challenging due to the high computational cost of existing detectors. Methods To address this, we propose UOD-YOLO, a lightweight real-time detection framework derived from YOLOv11n. The model integrates three key components: (1) a RepViT-based lightweight backbone for efficient feature representation; (2) a scale-sequence feature fusion (SSFF) module and a triple feature encoding (TFE) module to enhance multi-scale information extraction and fine-grained feature representation; and (3) a lightweight detection head combined with a channel–position attention mechanism (CPAM) to strengthen small-object recognition. We evaluate the model on three underwater benchmarks: RUOD, UTDAC2020, and DUO. Results Compared with YOLOv11n, UOD-YOLO improves mAP50 and mAP50–95 by 3.5% and 6.9%, respectively, while reducing the number of parameters and floating-point operations by 36.3% and 19.7%. The model achieves a real-time detection speed of 279.8 frames per second. Visualization experiments further confirm its robustness under varying illumination and depth conditions. Discussion The compact size and strong accuracy of UOD-YOLO make it suitable for deployment on embedded and edge devices in aquaculture monitoring, continuous underwater surveillance, and multi-sensor fusion systems. This work provides an efficient and reliable solution for real-time underwater organism detection in complex marine environments.