Spatial-Channel Attention-Based Class Activation Mapping for Interpreting CNN-Based Image Classification Models

Figure 1

ResNet-18 network: the feature maps of the middle layer ((b)∼(f), where conv1 denotes the first convolutional layer and conv2_x ∼ conv5_x denote the special designed convolutional modules in ResNet-18). The feature map visualization of the highest convolutional layer (g). (a) Original image; (b) conv1; (c) conv2_x; (d) conv3_x; (e) conv4_x; (f) conv5_x; (g) feature map visualization.

(a)

(b)

(c)

(d)

(e)

(f)

(g)