TABLE 1:

A summarization of image-to-image translation tasks and methods. $\{\cdot \}$ denotes the data used in a single training. $(\cdot ,\cdot )$ denotes the paired data between domains.

Image Translation Task	Representative Methods	Training Data Fixed Domains	Training Data Adding a Domain $X$	Inference
Paired dual-domain I2I	Pix2Pix [4], SPADE [5]	{($A$, $B$)}	-	$A\to B$
Unpaired dual-domain I2I	CycleGAN [1], UNIT [2]	{$A$, $B$}	-	$A\leftrightarrow B$
Unpaired multi-domain I2I	StarGAN [7], [8], DRIT++ [6]	{$A$, $B$, $C$}	{$A$, $B$, $C$, $X$}	$A$, $B$, $C$, $X\leftrightarrow A$, $B$, $C$, $X$
	Latent Anchor (Ours)	{$A$}, {$B$}, {$C$}	{$X$}	$A$, $B$, $C$, X ↔ A, $B$, $C$, $X$