The ocean contains a vast amount of rich and stable remote sensing data. Utilizing these data to realize intelligent real-time recognition of marine organisms is a critical task in marine remote sensing. Especially in complex seabed environments, where monitoring equipment is limited by computing power, oceanographers urgently require a detection algorithm with low computational complexity that can be widely deployed on various simple marine remote sensing devices. This is of great significance for marine remote sensing applications requiring real-time positioning of marine life, such as ecological protection and fishery management. This study proposes SEALNet, a novel fast detection network for seabed objects. The model integrates Mamba and YOLO principles to enable efficient lightweight benthic organism detection. For SEALNet’s neck, the original concatenation modules are improved, which efficiently aggregates feature layer information across backbone stages for cross-scale fusion. To further reduce the computational requirements of SEALNet, a new detection head module based on group normalization and shared convolution operations is designed. These improvements maintain a reasonable computational load while enhancing the precision of the object detection network. EUDD dataset tests indicate SEALNet’s performance: the detection precision achieves 90.6% (sea cucumbers), 91.6% (sea urchins), and 93.5% (scallops). Comparisons with mainstream models confirm its superiority in detecting benthic organisms. This work is expected to provide new insights and approaches for intelligent remote sensing and analysis in marine ranches.