search

Chrixtar / latentsplat

[ECCV 2024] Implementation of latentSplat: Autoencoding Variational Gaussians for Fast Generalizable 3D Reconstruction
https://geometric-rl.mpi-inf.mpg.de/latentsplat/
MIT License
146 stars 3 forks source link

how can I train on re10k? #6

Closed BroenLin closed 5 months ago

BroenLin commented 5 months ago

Is there a corresponding command?

Chrixtar commented 5 months ago

Hi BroenLin,

the explanations in the Readme just use co3d_hydrant as an example. The command for training on re10k is (referring to the config/experiment/re10k.yaml):

python3 -m src.main +experiment=re10k

For evaluation of extrapolation on re10k, the command would be:

python3 -m src.main +experiment=re10k mode=test dataset/view_sampler=evaluation dataset.view_sampler.index_path=assets/evaluation_index/re10k_extra.json checkpointing.load=checkpoints/re10k.ckpt

I recommend checking out the hydra documentation, as we use it for our experiment configurations.

Best Chris