The paper presents an improved YOLOv8-based algorithm, called Ship-Fire Net, for detecting fires and smoke in ships. The authors address the limitations of conventional detection systems, which are often ineffective due to distance constraints and ship motion. They propose a lightweight algorithm that combines YOLOv8n with GhostnetV2-C2F for long-range attention and spatial-channel reconstruction convolution (SCConv) to reduce redundant features. The algorithm also incorporates omni-dimensional dynamic convolution for multi-dimensional attention. The dataset includes over 4000 images of fires inside ship rooms and on board, manually labeled to ensure quality. The proposed method achieves a precision and recall of over 0.93 for fire and smoke detection, with an mAP@0.5 of about 0.9. It also has fewer parameters and lower computational costs compared to the original YOLOv8n, achieving a detection speed of 286 FPS, suitable for real-time ship fire monitoring. The paper discusses the development of the dataset, experimental setup, and evaluation metrics, and compares the proposed method with other advanced algorithms, demonstrating its superior performance in terms of accuracy and speed.The paper presents an improved YOLOv8-based algorithm, called Ship-Fire Net, for detecting fires and smoke in ships. The authors address the limitations of conventional detection systems, which are often ineffective due to distance constraints and ship motion. They propose a lightweight algorithm that combines YOLOv8n with GhostnetV2-C2F for long-range attention and spatial-channel reconstruction convolution (SCConv) to reduce redundant features. The algorithm also incorporates omni-dimensional dynamic convolution for multi-dimensional attention. The dataset includes over 4000 images of fires inside ship rooms and on board, manually labeled to ensure quality. The proposed method achieves a precision and recall of over 0.93 for fire and smoke detection, with an mAP@0.5 of about 0.9. It also has fewer parameters and lower computational costs compared to the original YOLOv8n, achieving a detection speed of 286 FPS, suitable for real-time ship fire monitoring. The paper discusses the development of the dataset, experimental setup, and evaluation metrics, and compares the proposed method with other advanced algorithms, demonstrating its superior performance in terms of accuracy and speed.