This paper proposes a lightweight YOLOv8 model, named UW-YOLOv8, for real-time underwater object detection. The model optimizes YOLOv8 to better suit underwater environments by replacing its original backbone with a lightweight FasterNet module, which reduces computation and improves performance. The BiFPN is modified to be faster by removing unnecessary layers and changing the fusion method. Additionally, a lightweight-C2f structure is proposed by replacing the standard convolution and bottleneck module with GSConv and partial convolution, resulting in a lighter and faster block. Experiments on three underwater datasets (RUOD, UTDAC2020, URPC2022) show that the proposed method achieves mAP50 of 86.8%, 84.3%, and 84.7%, respectively, with a speed of 156 FPS on NVIDIA A30 GPUs. Compared to YOLOv8, the model size is reduced by 24% on average, and mAP accuracy is improved on all three datasets. Underwater object detection is essential for exploring and protecting marine ecosystems and monitoring fish populations in aquaculture. However, underwater environments pose challenges such as low image contrast, blurred targets, and excessive noise due to limited light availability and optical scattering. Existing methods include two-stage and one-stage approaches. Two-stage methods, like Faster-RCNN, are accurate but slow and have large model sizes. One-stage methods, such as SSD, YOLOv5, and YOLOv7, are faster but may struggle with small targets and object direction sensitivity. Recent improvements include multi-directional edge detection, feature enhancement, and attention mechanisms to enhance detection accuracy for underwater objects. Lightweight networks, such as MobileNetV3 and ShuffleNet, are designed for efficiency and are increasingly used in detection models. The proposed UW-YOLOv8 integrates a lightweight FasterNet backbone and a modified BiFPN to achieve a balance between speed, accuracy, and model size, making it suitable for real-time underwater object detection.This paper proposes a lightweight YOLOv8 model, named UW-YOLOv8, for real-time underwater object detection. The model optimizes YOLOv8 to better suit underwater environments by replacing its original backbone with a lightweight FasterNet module, which reduces computation and improves performance. The BiFPN is modified to be faster by removing unnecessary layers and changing the fusion method. Additionally, a lightweight-C2f structure is proposed by replacing the standard convolution and bottleneck module with GSConv and partial convolution, resulting in a lighter and faster block. Experiments on three underwater datasets (RUOD, UTDAC2020, URPC2022) show that the proposed method achieves mAP50 of 86.8%, 84.3%, and 84.7%, respectively, with a speed of 156 FPS on NVIDIA A30 GPUs. Compared to YOLOv8, the model size is reduced by 24% on average, and mAP accuracy is improved on all three datasets. Underwater object detection is essential for exploring and protecting marine ecosystems and monitoring fish populations in aquaculture. However, underwater environments pose challenges such as low image contrast, blurred targets, and excessive noise due to limited light availability and optical scattering. Existing methods include two-stage and one-stage approaches. Two-stage methods, like Faster-RCNN, are accurate but slow and have large model sizes. One-stage methods, such as SSD, YOLOv5, and YOLOv7, are faster but may struggle with small targets and object direction sensitivity. Recent improvements include multi-directional edge detection, feature enhancement, and attention mechanisms to enhance detection accuracy for underwater objects. Lightweight networks, such as MobileNetV3 and ShuffleNet, are designed for efficiency and are increasingly used in detection models. The proposed UW-YOLOv8 integrates a lightweight FasterNet backbone and a modified BiFPN to achieve a balance between speed, accuracy, and model size, making it suitable for real-time underwater object detection.