MSCB-YOLOv8: Multi-Scale Cigarette Butt Small Object Detection Model with Channel Prior Convolutional Attention

In power-related environments, casually discarded cigarette butts pose a significant fire hazard. Therefore, the development of accurate cigarette butt detection methods is crucial. Traditional cigarette butt detection methods suffer from high false alarm rates and poor adaptability due to the complexity of cigarette butt shapes, textures, and environmental conditions. To address these challenges, a multi-scale cigarette butt small object detection model based on YOLOv8n enhancements is proposed. Firstly, a novel Channel Prior Convolutional Attention (CPCA) module is introduced to dynamically allocate attention weights in both channel and spatial dimensions, utilizing multi-scale depthwise separable convolution modules to construct spatial attention. This module aims to improve small object detection performance effectively while reducing computational costs by better capturing key information from each channel in the image. Secondly, by introducing a small object detection layer, the model can learn and analyze image features, identify small objects, and label their positions and categories. Experimental results on the newly constructed cigarette butt dataset demonstrate the effectiveness of the proposed model. Our model achieves an mAP of 86.20%, which is approximately 2.4% higher than that of the baseline YOLOv8n network. This method notably enhances the precision of detecting cigarette butt small objects, lowers false alarm rates, and showcases outstanding suitability in real-world power safety scenarios.