Closed caiyongqi closed 1 year ago
https://github.com/Boese0601/RC-MVSNet/blob/c8dfe7c6c3ebadd47d70087aac776a268a478fae/train_rcmvsnet.py#L376 Hi, I found that the MVSNeRF module only seems to affect the gradient of the image encoder, how does it improve the depth estimation of CascadeMVSNet?
We use a share weight encoder so that the backbone depth estimation network is affected by the joint trained gradient from MVSNeRF module.
https://github.com/Boese0601/RC-MVSNet/blob/c8dfe7c6c3ebadd47d70087aac776a268a478fae/train_rcmvsnet.py#L376 Hi, I found that the MVSNeRF module only seems to affect the gradient of the image encoder, how does it improve the depth estimation of CascadeMVSNet?