This paper proposes an improved YOLOv8 model, named Our-v8, for the detection of coal and gangue. The model is designed to address the challenges of lightweight and real-time detection in coal sorting. The dataset used includes images of coal and gangue under two different lighting environments, with images enhanced using the Laplacian algorithm to improve training data quality. The CBAM attention mechanism is introduced to prioritize crucial features, enhancing feature extraction. The EIOU loss function is added to refine box regression, improving detection accuracy. The model achieves a mean average precision (mAP) of 99.5% in a halogen lamp lighting environment, with a model size of only 22.1 MB, FLOPs of 29.7, and parameters of 12.8. Our-v8 provides accurate location information for coal and gangue, making it suitable for real-time coal sorting applications. The model outperforms existing YOLO series models in terms of accuracy and efficiency, with higher detection speed and lower computational requirements. The study also compares Our-v8 with other advanced models and existing research, demonstrating its superior performance in coal and gangue detection. The results show that Our-v8 can effectively detect coal and gangue under different lighting conditions, with high accuracy and speed, making it a promising solution for real-time coal sorting.This paper proposes an improved YOLOv8 model, named Our-v8, for the detection of coal and gangue. The model is designed to address the challenges of lightweight and real-time detection in coal sorting. The dataset used includes images of coal and gangue under two different lighting environments, with images enhanced using the Laplacian algorithm to improve training data quality. The CBAM attention mechanism is introduced to prioritize crucial features, enhancing feature extraction. The EIOU loss function is added to refine box regression, improving detection accuracy. The model achieves a mean average precision (mAP) of 99.5% in a halogen lamp lighting environment, with a model size of only 22.1 MB, FLOPs of 29.7, and parameters of 12.8. Our-v8 provides accurate location information for coal and gangue, making it suitable for real-time coal sorting applications. The model outperforms existing YOLO series models in terms of accuracy and efficiency, with higher detection speed and lower computational requirements. The study also compares Our-v8 with other advanced models and existing research, demonstrating its superior performance in coal and gangue detection. The results show that Our-v8 can effectively detect coal and gangue under different lighting conditions, with high accuracy and speed, making it a promising solution for real-time coal sorting.