/RMVSNET_Unofficial_Torch

Unofficial RMVSNET implementation

Primary LanguagePython

RMVSNET_Unofficial_Torch

Warning

Unfinished ! training loss high , possibly something went wrong with the code.

Currently using models.convgru for cost regularization

Only implement forward cost regularization

Update

2023/4/14 training loss goes down , tested within 1 epoch

2023/5/10 move to unstable branch for more updates

Description

Unofficial RMVSNET implementation .

Code mostly borrowed from

  1. MVSNet_pytorch
  2. RMVSNet-Pytorch
  3. loss function from AA-RMVSNet

File Tree

root
  |
  |___datasets
  |
  |___evaluations
  |
  |___lists
  |
  |___models

Run

Modify data root and other parameters in train.sh

run sh train.sh