Skip to main content
. 2020 Sep 2;48(3):721–728. doi: 10.1007/s00259-020-05006-3

Fig. 1.

Fig. 1

Architecture of the CNN. Each convolution block consists of two convolution layers, batch normalization and two ReLu activation functions. Max pooling is performed to down sample the data. Using a sigmoid function weights (Wn) are added to the nodes generated by the GAP layer. Network activation mapping is applied for object localization