Table 6.

Performance comparison between the improved Mask R-CNN backbone and the original one on the Grape-B test set.

Backbone	ResNet50-FPN		ResNet50-FPN-ED
	Box	Seg	Box	Seg
$AP$	58.80	57.40	59.70	58.80
$A{P}_{IoU=0.5}$	85.50	87.00	86.50	87.20
$A{P}_{IoU=0.75}$	64.60	62.90	64.80	66.10
$A{R}_{\max =1}$	7.30	7.10	7.70	7.50
$A{R}_{\max =10}$	54.20	52.60	54.90	53.40
$A{R}_{\max =100}$	67.20	64.40	69.10	66.10

The best one is highlighted in bold.