WCD-YOLO: A waste classification detection model

To solve the problem of intelligent waste classification, a new waste classification detection model named WCD-YOLO (Waste Classification Detection -You Only Look Once) is proposed. We first optimize the backbone network of YOLOv10 by developing an MCA module to heighten the overall model's feature extraction capability and refine the model's precision. At the same time, a more powerful feature extraction module, FNC2f, with high efficiency and multi-scale characteristics, is created to enrich the network's feature extraction and meet the needs of high-precision target recognition. A brand-new FNC2f-BiFPN feature pyramid network structure is also designed, strengthening the detection ability of the waste with insufficient feature capture. Finally, we use Inner-CIoU as the loss function of the WCD-YOLO model and control the scale of the auxiliary boundary. Experimental results show that the WCD-YOLO model developed in this research has better precision on the self-built dataset at two IoU thresholds than other models, with mAP₅₀ reaching 95.8%, an increase of 1.6% over the original model, and mAP_50:95 reaching 74.0%, an increase of 2.6%. The model parameters are only 7.2MB, and the GFLOPs are 8.5G. The proposed model is characterized by low consumption and high precision in waste recognition and classification, providing a reference for future academic research and engineering practice.