About the depth_loss, if using it, will the result get better or worse ?

[MaxChanger]

Hello, thanks for the amazing work. I found some discussion about depth_loss in code #L129 and issue https://github.com/apchenstu/mvsnerf/issues/34. And, I know that this loss is only used when --with_depth_loss, the author does not seem to use this depth loss in paper and code usage, but only image-level L2 loss.

I would like to know, if the supervision of this depth loss is added, will the quality of image rendering be improved? Has the author conducted relevant experiments, or is this loss_func a legacy code of MVSNet?

Thanks.