Table 2.

Dimensional change table for DAM Module.

Step	Operation	Input dimension	Output dimension	Descriptions
1	convolutional encoding	[m, c, w, h]	[m, 1, w, h]	Generating Attention Maps
2	Weighting support features with attention maps	[m, 1, w, h]	[m, c, w, h]	Support for feature weighting
3	Flatten the feature map spatial dimension	[m, c, w, h] [n, c,w, h]	[m, c,l] [n, c,l]	Calculate the similarity between the query and the supporting features and generate the similarity matrix
4	Softmax for normalization	[m, c,l] [n, c,l]	[m, n]	Get the attention map
5	Weighting Value	[m, n]	[m, c, w, h]	Fusing support features to query features
6	Integration support features	[m, c, w, h]	[c, w,h]	Integration of the two branches of attention