Attention-Based Image-to-Video Translation for Synthesizing Facial Expression Using GAN

Table 1

List of experimental classes conducted.

Notation	Experiment	Dataset used
AffineGAN	Baseline work	MUG and local facial expression dataset
MAEGAN	Expression intensity inferred as stated, boundary image used as an informative region for the local discriminator and MAE loss used to train the generator and the discriminators	MUG and local facial expression dataset
MAEGAN + SA	With the addition of the self-attention layer and with local discriminator	MUG and local facial expression dataset