Algorithm 1 The implementation of local spatial feature suppression.

Here, the feature map is $X_l=\{x_1,x_2,\cdots ,x_m\}$, $r\sim Bernoulli\left(p\right)$ has p probability of being 1, c is the number of network channels, and $ct$ is a random integer to control the number of loops.

for$l=1,\cdots ,c$do

Initialize r;

if $r==1$ then

Initialize $ct$;

for $i=1,\cdots ,ct$ do

${\tilde{X}}_l=P^{\prime }\circ X_l$;

else

${\tilde{X}}_l=X_l$;