Evaluation metric when loss is MSE

If I have 3 channel images (RGB) as inputs and say the network is prediction 5 channels of values where all the values range from 0.0-1.0, I believe a good choice of loss would be MSE since its a regression problem.

However, I am having a hard time on deciding a good evaluation metric for this type of a task. What can be a good choice? From practice I believe using the same function (MSE) will not be a good way to evaluate right?

1 Like