Skip to main content

Table 2 Experimental results for different feature fusion methods

From: Deep learning for predicting refractive error from multiple photorefraction images

Methods

MAE (D)

Accuracy (%)

MAE (D)

Accuracy (%)

MAE (D)

Accuracy (%)

Spherical component

Cylindrical component

Spherical equivalent

Addition

0.2818

83.06

0.1210

96.39

0.2593

85.98

Concatenation

0.3663

74.02

0.1590

96.59

0.3196

78.19

AFF

0.2443

87.05

0.0761

96.40

0.2109

89.00

LSTM

0.1740

89.50

0.0702

96.70

0.1835

89.38

  1. Bold values indicate the best results