CA-XTree: Age Estimation of Grouped Gradient Regression Tree with Local Channel Attention

Table 5

Ablation experiments on CACD. CA is channel attention.

Baseline	Setting	MAE (CACD)	CS(3) (CACD) (%)
ResNet-50	c = 2_CA_tree = 7	4.639	62.50
ResNet-50	c = 4_CA_tree = 7	4.578	63.04
ResNet-50	c = 8_CA_tree = 7	4.593	62.58
ResNet-50	c = 16_CA_tree = 7	4.575	62.23
ResNet-50	c = 32_CA_tree = 7	4.598	62.48

“c” indicates the number of groups. CA stands for local channel attention.