yolov8vsyolov10

YOLOv10 introduces a dual-head architecture for NMS-free training and efficiency-accuracy driven model design. It combines one-to-one and one-to-many label assignments to improve performance without extra computation. YOLOv10 uses lightweight classification