jkulhanek / tetra-nerf

Official implementation for Tetra-NeRF paper - NeRF represented as triangulation of input point cloud.
https://jkulhanek.com/tetra-nerf
MIT License
278 stars 14 forks source link

Show training progress of tetra-nerf like nerfacto not no wandb #8

Closed ymd8bit closed 1 year ago

ymd8bit commented 1 year ago

I recently started to use this brilliant project, and I'm looking for the way to see the training progress with visualization like nerfacto on nerfstudio. Is it possible to visualize tetra-nerf rendering result during ns-train by viewer.nerf.studio?

jkulhanek commented 1 year ago

Hi, thanks! It should be possible. Just specify --vis viewer flag to enable. You can explore the trained checkpoints first by using `ns-viewer --load-config ...path-to-downloaded-checkpoint..."

ymd8bit commented 1 year ago

I tried that way, but the vis flag seems not to be supported with tetra-nerf. I pasted the error message below. point cloud data pointnerfj-0.5.th was made from the one point-nerf provided.

ns-train tetra-nerf --pipeline.model.tetrahedra-path data/blender/lego/pointnerf-0.5.th blender-data --data /data/nerf_synthetic/lego --vis viewer
usage: ns-train [-h]
                {nerfacto,nerfacto-big,depth-nerfacto,volinga,instant-ngp,instant-ngp-bounded,mipnerf,semantic-nerfw,vanilla-nerf,tensorf,dnerf,phototourism,nerfplayer-nerfacto,nerfplayer-ngp,neus,neus-
facto,tetra-nerf,tetra-nerf-original}

ns-train: error: unrecognized arguments: --vis viewer
jkulhanek commented 1 year ago

The vis flag has to be added before blender-data.

jkulhanek commented 1 year ago

Please reopen if it doesn't work.