About the GPU memory cost in DTU training.

[Haonan-DONG]

Hi, thanks for the excellent work for unsupervised learning based MVS. I have one simple question since the code is not published.

How is the GPU memory cost in every stage if the two stages are separately trained on DTU_train datasets?

Thanks.

[ToughStoneX]

Sorry for the late response. As discussed in the paper, the two stages (pretraining stage and post-training stage) are conducted sequentially. They share the same backbone of CascadeMVSNet. And the GPU memory cost is the same as the default configuration of CascadeMVSNet. It is no more than 10G, if I remember it right.