Skip to main content
. 2021 Mar 31;9:e11182. doi: 10.7717/peerj.11182

Table 3. Performance measures of test datasets obtained via TL-CNN pipelines.

Class Precision Recall F1-score Specificity CA
Performance measures obtained through inceptionV3 pipeline
Left winking 0 1.00 1.00 1.00 1.00 1.00
Right winking 1 1.00 1.00 1.00 1.00
No winking 2 1.00 1.00 1.00 1.00
Performance measures obtained through inception ResNetV2 pipeline
Left winking 0 0.86 1.00 0.92 0.14 0.94
Right winking 1 1.00 1.00 1.00 0.00
No winking 2 1.00 0.83 0.91 0.17
Performance measures obtained through Xception pipeline
Left winking 0 0.50 1.00 0.67 0.50 0.67
Right winking 1 1.00 1.00 1.00 0.00
No winking 2 0.00 0.00 0.00 1.00
Performance measures obtained through MobileNet pipeline
Left winking 0 1.00 0.33 0.50 0.77 0.78
Right winking 1 1.00 1.00 1.00 0.00
No winking 2 0.60 1.00 0.75 0.40