Skip to main content
. Author manuscript; available in PMC: 2023 May 1.
Published in final edited form as: Med Image Anal. 2022 Mar 22;78:102424. doi: 10.1016/j.media.2022.102424

Figure 4:

Figure 4:

Detailed network architecture of FedReplay when ResNet34 (He et al., 2016) is used as the backbone network. We omit ReLu, batch normalization and max pooling layers that follow each convolution layer. We split the ResNet34 (He et al., 2016) into two parts, where the first part was used as the auxiliary encoder network, and the remaining part was used as the primary classifier network.