Skip to main content
. 2022 Jun 19;22(12):4633. doi: 10.3390/s22124633

Table 1.

Comparison of input features and type of methods.

Methods Input Features Type Limitation
Light-CNN [19] Lower + Upper Face-based Mainly rely on Lower
eXnet [15] Lower + Upper Face-based Mainly rely on Lower
Pre-trained CNN [19] Lower + Upper Face-based Mainly rely on Lower
PG-CNN [14] Lower + Upper Face-based Mainly rely on Lower
DLP-CNN [12] Lower + Upper Face-based Mainly rely on Lower
gACNN [13] Lower + Upper Face-based Mainly rely on Lower
RASnet [11] Lower + Upper Face-based Mainly rely on Lower
DeRL [16] Lower + Upper Face-based Mainly rely on Lower
ResiDen [20] Lower + Upper Face-based Mainly rely on Lower
SHCNN [21] Lower + Upper Face-based Mainly rely on Lower
ResNet-PL [22] Lower + Upper Face-based Mainly rely on Lower
RAN [23] Lower + Upper Face-based Mainly rely on Lower
SCN [24] Lower + Upper Face-based Mainly rely on Lower
DACL [25] Lower + Upper Face-based Mainly rely on Lower
ARM [26] Lower + Upper Face-based Mainly rely on Lower
RTFER [11] Lower + Upper Constituent-based Mainly rely on Lower
DenseNet [27] Lower + Upper Constituent-based Mainly rely on Lower
OADN [28] Lower + Upper Constituent-based Mainly rely on Lower
Proposed Method Upper Constituent-based Rely only on Upper