Skip to main content
. 2022 Aug 30;11(17):5102. doi: 10.3390/jcm11175102

Figure 4.

Figure 4

Overview of transfer learning process. The weights obtained on the ImageNet dataset are transferred to the convolution layers. The weights in the fully connected part are retrained. After optimization, the four deep learning models have two 128 node dense layers and one dropout layer with a 0.2 ratio as fully connected layers.