GAN	Generative Adversarial Network
CNN	Convolutional Neural Network
FCN	Fully Convolutional Network
SE	Squeeze and Excitation block
DC	Dilated Convolution
TTUR	Two Time-scale Update Rules
Notations
$l_c$	Occluded image
$l_{gt}$	Ground truth image
$l_{pre\_mask}$	Generated binary mask
$l_m$	Noise-free binary mask
$l_c$	$\mathrm{Concatenated} \mathrm{input} \mathrm{of} \mathrm{occluded} \mathrm{image} l_c$ $\mathrm{and} \mathrm{generated} \mathrm{binary} \mathrm{mask} l_{pre\_mask}$
$I_o^i$	Initially generated de-occluded facial image
$I_o^f$	$\mathrm{Concatenated} \mathrm{input} \mathrm{of} \mathrm{occluded} \mathrm{image} l_c$ $, \mathrm{generated} \mathrm{binary} \mathrm{mask} l_{pre\_mask} \mathrm{and}$ $\mathrm{initially} \mathrm{generated} \mathrm{de}-\mathrm{occluded} \mathrm{facial} \mathrm{image} I_o^i$
$I_i^f$	Finally generated de-occluded facial image