Closed hardikdava closed 4 months ago
Of course. Here is an example config file:
model:
gaussian:
optimization:
densify_grad_threshold: 0.0006
renderer: internal.renderers.gsplat_renderer.GSPlatRenderer
absgrad: true
light_gaussian:
prune_decay: 0.6
prune_percent: 0.6
prune_steps:
- 16_000
- 24_000
@yzslab Thanks for the config. It worked! what do you think about integrating your gsplat version of light gaussian to nerfstudio gsplat?
Thanks for the config. It worked!
Glad to hear that.
what do you think about integrating your gsplat version of light gaussian to nerfstudio gsplat?
My modifications bring some code redundancies to gsplat. https://github.com/yzslab/gsplat/commit/747f55b003450e6a690b41f6cee0e0548601e1b3 https://github.com/yzslab/gsplat/commit/ca617c5b8012773dbd53442228f65b789a62094f
May be someday when I am willing to write a better version. (Don't expect too much.)
Hello, is it possible to train a model with a combination of gsplat absgs and light gaussian? Thanks