Garbage detection model based on improved YOLOv8n

In order to address missed detections, false positives, and low precision caused by dense targets, occlusions, complex backgrounds, and multi-scale garbage objects in challenging scenes, this paper proposes improvements to a YOLOv8n-based garbage detection model. We introduce Receptive-Field Attention Convolution (RFAConv) to replace the traditional convolutions in the backbone, optimizing the model’s ability to process features at different receptive field scales and thereby enhancing its feature extraction capability in complex environments. Additionally, we employ the Adaptive Spatial Feature Fusion (ASFF) mechanism in the detection head to combine semantic features from high-level, large-receptive-field maps with detailed information from lowlevel, high-resolution maps. This integration enables the model to better handle multi-scale garbage targets and improves both detection precision and computational efficiency. Experimental results demonstrate that the improved model achieves a 1.5% increase in mAP@ 0.5 compared to the baseline, without a significant increase in parameters or computational overhead; precision and recall also show improvement. Overall, the enhanced model shows substantially better detection performance in complex environments and more effectively completes garbage detection tasks across diverse scenarios.