Multi-element Integration and Large Kernel Attention-enhanced Aerial Target Detection

LH Net-YOLO, a high-res aerial target detection method, addresses low-res drone image issues. It uses HybridGAP-Net for better resolution and multi-scale adaptability, SCConv for lightweighting, Large Kernel Attention for small targets,and MPDIoU loss function for reduced missed detections. This method is applied on the VisDrone2019 and DOTA1.0 datasets [email protected] P, R [email protected] On 0.95, it performs better than the YOLOv8n model and also shows improvement at other scales of YOLOv8, confirming its effectiveness.