Why do you add MSE loss to L1 loss?

[miquel-espinosa]

In LPIPSWithDiscriminator, why do you add MSE loss to the L1 loss?
In the original code they do NOT add MSE loss to the L1 loss.

I really appreciate your answer. Thanks,

[GuHuangAI]

Generally speaking, L1 preserves the structure, and L2 ensures overall consistency.