Skip to main content
. 2021 Jul 1;21(13):4535. doi: 10.3390/s21134535

Figure 3.

Figure 3

Architecture of the applied convolutional neural network. The size of the dense layers was found during the hyperband search and differs for the joint angle and joint moment prediction. For the pretrained model, the convolutional and pooling layers were set to be non-trainable.