YOLOv8-CLS: an improved model for floating waste detection on water surfaces

To address the challenges of complex background interference and small object detection in floating garbage detection, this paper proposes an improved algorithm based on YOLOv8, named YOLOv8-CLS. A probability-based data augmentation strategy is employed, incorporating geometric, optical, and environmental transformations to enhance model adaptability. First, the C2f_MultiOGA structure is introduced into the neck network to improve small object detection performance. Then, the detection head is replaced with LSCD (Local Spatial Context Detection), and ShapeIoU is integrated as the loss function to optimize bounding box regression and improve localization accuracy. Experimental results show that the proposed method achieves an mAP@50 of 93.8% and an mAP@50-95 of 59.7%, outperforming the original YOLOv8 by 0.5% and 1%, respectively, demonstrating the effectiveness of the proposed approach.