Figure 2
The diagram illustrates the ResNet50-CBAM model architecture. It begins with preprocessing, followed by ResNet50 base layers. The output feature F is processed through a channel attention module, which includes average pooling, max pooling, and shared dense layers, producing channel attention. This refined feature is then passed through a spatial attention module, involving convolutional layers and spatial attention, resulting in the CBAM feature. The final output is obtained after global average pooling.The ResNet50-CBAM model architecture