yucornetto/MGMatting

About the metrics in training and testing

crystraldo opened this issue · 0 comments

Hi, I found that the sad/mse tested during training is different from the sad/mse tested with inference, what is the specific reason? Is it because of the dist problem?