Algorithm 1: Modified Algorithm Training Loop Pseudocode
1:	Draw a minibatch of samples {XAB(1),..., xAB(m)} from domain X
2:	Draw a minibatch of samples {Y(1),..., y(m)} from domain Y
3:	Compute the discriminator loss on real images: ${\mathcal{I}}_{real}^{\left(D\right)}=\frac{1}{m}{{\displaystyle \sum }}_{i=1}^m{\left(D_{\mathrm{A}}\left(A^{\left(i\right)}\right)-1\right)}^2$ + $\frac{1}{n}{{\displaystyle \sum }}_{i=1}^n{\left(D_B\left(B^{\left(j\right)}\right)-1\right)}^2$
4:	Compute the discriminator loss on fake images:     ${\mathcal{I}}_{fake}^{\left(D\right)}=\frac{1}{m}{{\displaystyle \sum }}_{i=1}^m{\left(D_{\mathrm{B}}\left(G_{\mathrm{A}\to \mathrm{B}}\left(A^{\left(i\right)} \right)\right)\right)}^2$ + $\frac{1}{n}{{\displaystyle \sum }}_{j=1}^n{\left(D_{\mathrm{A}}\left(G_{\mathrm{B}\to \mathrm{A}}\left(B^{\left(j\right)} \right)\right)\right)}^2$
5:	Update the PatchGAN discriminator
6:	Apply Attention to the Generator         $q_{att}^l={\psi }^T\left({\sigma }_1\left(W_x^T{x}_i^l+W_g^T{g}_i+b_g\right)\right)+b_{\psi }$)             ${\alpha }_i^l={\sigma }_2\left(q_{att}^l\left(x_i^l,g_i; {\Theta }_{att}\right)\right)$
7:	Compute the B → A generator loss:          ${\mathcal{J}}^{\left(G_{B\to A}\right)}=\frac{1}{n}{\displaystyle \sum }_{j=1}^n{D}_A\left(G_{B\to A}\left(B^{\left(j\right)}\right)\right)-1{)}^2+{\mathcal{J}}^{B\to A}$
8:	Compute the A → B generator loss:          ${\mathcal{J}}^{\left(G_{A\to B}\right)}=\frac{1}{m}{\displaystyle \sum }_{i=1}^m{D}_B\left(G_{A\to B}\left(A^{\left(i\right)}\right)\right)-1{)}^2+{\mathcal{J}}^{A\to B}$
9:	Update the generator