Skip to main content
. 2022 Oct 18;19(20):13491. doi: 10.3390/ijerph192013491

Table 3.

Performance of deep learning models (EfficientNetB4, ResNet50, and ECA-Net50) with pose estimator input trained using datasets with no blanket, all blankets, and augmented data.

Outcome Dataset Deep Learning Models with Pose Estimator
EfficientNetB4 ResNet50 ECA-Net50
Accuracy No blanket data 91.1% 91.7% 92.3%
All blanket data 91.4% 89.3% 90.9%
Augmented data 89.7% 91.1% 91.5%
F1-score No blanket data 91.1% 91.6% 92.2%
All blanket data 91.9% 88.9% 91.7%
Augmented data 90.8% 91.3% 92.2%