janusch

Results 50 issues of janusch

Replace it. We want free software. And we want to control our code to easily improve it. Provide tests.

help wanted
good first issue
intermediate

Ply saving based on open3D. That's a quick way to generate ply files. I think this is a highly requested feature. Should work with default and mcmc densification strategies.

This is a draft implementation of scaffold-gs. It is still very raw, but I wanted to get the base running as quickly as possible. The simple_trainer.py is already functional, and...

- Should run in its own gui thread. - It should reuse the rasterizer - Should allow for simple controls like fps or orbiting.

enhancement
help wanted

Compare my pr for saving: https://github.com/nerfstudio-project/gsplat/pull/427 Another aspect: I really don't think we need to save the initial point cloud as ply as done in the original code. That part...

help wanted
good first issue

Code (MIT) can be copied over. Credits to the original author are required. https://github.com/rahul-goel/fused-ssim

enhancement
help wanted
good first issue

Implement mcmc for the densification as strategy. See https://github.com/nerfstudio-project/gsplat/blob/main/gsplat/strategy/mcmc.py https://arxiv.org/abs/2404.09591 Depends on [Issue](https://github.com/MrNeRF/gaussian-splatting-cuda/issues/60)

enhancement
help wanted

Factor out the densification into a strategy so that we can replace the method similar to gpslat https://github.com/nerfstudio-project/gsplat/tree/main/gsplat/strategy

help wanted
good first issue

This PR implements [this paper](https://arxiv.org/abs/2503.19232), which improves background reconstruction while preserving foreground fidelity compared to training without homogeneous coordinates. I’d appreciate community support to verify this, but based on my...

**Build a system that automatically finds optimal hyperparameters for each scene.** **Challenge.** Different scenes need different hyperparameters. Create an automatic optimization system (RL-based or other approaches) that discovers the best...