# change predict score based on sigma

[Rickyyy-zh]

Hello, here I find that the prediction score has been weighted by depth. I assume the reason is with larger depth, the prediction is more uncertain. but I didn't notice this method mentioned in your paper？Is this operation important for results or training?

 # change predict score based on sigma
        sigma = depth_pred[:, :, 1]
        sigma = torch.exp(-sigma)
        batch_scores *= sigma
        batch_scores = batch_scores[..., None]