Sensors, Год журнала: 2024, Номер 25(1), С. 181 - 181

Опубликована: Дек. 31, 2024

Recently, computer vision methods have been widely applied to agricultural tasks, such as robotic harvesting. In particular, fruit harvesting robots often rely on object detection or segmentation identify and localize target fruits. During the model selection process for detection, average precision (AP) score typically provides de facto standard. However, AP is not intuitive determining which most efficient It based intersection-over-union (IoU) of bounding boxes, reflects only regional overlap. IoU alone cannot reliably predict success gripping, identical scores may yield different results depending overlapping shape boxes. this paper, we propose a novel evaluation metric To assess gripping success, our uses center coordinates boxes margin hyperparameter that accounts gripper’s specifications. We conducted about popular models peach apple datasets. The experimental showed proposed much more helpful in interpreting performance data.

Язык: Английский