This paper proposes Ship-Fire Net, an improved YOLOv8 algorithm for ship fire detection. Ship fires pose significant risks to safety, cargo, and the environment, necessitating prompt detection. Conventional systems face challenges due to distance and motion constraints, while deep learning offers potential solutions but is hindered by computational complexity. To address this, the authors developed a lightweight algorithm based on YOLOv8n, incorporating GhostnetV2-C2F for long-range attention, SCConv for feature reduction, and omni-dimensional dynamic convolution for multi-dimensional attention. These improvements result in a more accurate and efficient model, achieving precision and recall above 0.93 for fire and smoke detection, with mAP@0.5 of 0.9. Ship-Fire Net also has fewer parameters and lower FLOPs than the original YOLOv8n, enabling faster detection (286 FPS). The dataset includes over 4000 images of fires inside and outside ships, with labels for fire and smoke. The model was tested against other algorithms, showing superior performance in accuracy and speed. The study highlights the importance of detecting both fire and smoke, as smoke is a critical early indicator of fire. Ship-Fire Net is designed for real-time monitoring, with a focus on reducing computational complexity and improving detection efficiency. The results demonstrate that Ship-Fire Net is a lightweight, accurate, and efficient solution for ship fire detection.This paper proposes Ship-Fire Net, an improved YOLOv8 algorithm for ship fire detection. Ship fires pose significant risks to safety, cargo, and the environment, necessitating prompt detection. Conventional systems face challenges due to distance and motion constraints, while deep learning offers potential solutions but is hindered by computational complexity. To address this, the authors developed a lightweight algorithm based on YOLOv8n, incorporating GhostnetV2-C2F for long-range attention, SCConv for feature reduction, and omni-dimensional dynamic convolution for multi-dimensional attention. These improvements result in a more accurate and efficient model, achieving precision and recall above 0.93 for fire and smoke detection, with mAP@0.5 of 0.9. Ship-Fire Net also has fewer parameters and lower FLOPs than the original YOLOv8n, enabling faster detection (286 FPS). The dataset includes over 4000 images of fires inside and outside ships, with labels for fire and smoke. The model was tested against other algorithms, showing superior performance in accuracy and speed. The study highlights the importance of detecting both fire and smoke, as smoke is a critical early indicator of fire. Ship-Fire Net is designed for real-time monitoring, with a focus on reducing computational complexity and improving detection efficiency. The results demonstrate that Ship-Fire Net is a lightweight, accurate, and efficient solution for ship fire detection.