CompVis / CompVis/EDGS

`points3D.bin` - Making EDGS easier to use with other 3DGS methods

Open
#15 2 comments 4 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Jupyter Notebook
Stars
742
Forks
59
PR merge metrics
No merged PRs in 30d

Description

What about exporting the dense point cloud as a points3D.bin file instead of in memory overriding the GaussianModel's splats? I think this would turn EDGS into an easier module that can be slotted into any 3DGS pipeline similar to bilagrid's success.

This suggestion is based on my understanding of the code that the key code implementing EDGS improvement on 3DGS is in init_gaussians_with_corr and after creating the dense point cloud the code uses it to initialize more gaussians and proceeds to training. This all happens in memory and so there's no intermediate output that could just be passed as an input to another 3DGS method. Most 3dgs methods will either use a random initialization or the sparse points3D.bin point cloud generated by colmap. If EDGS outputted points3D.bin it could be an easy addition to any other 3DGS method.

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by tracing init_gaussians_with_corr and the code that creates the dense point cloud before it overrides the GaussianModel's splats. Check how COLMAP's points3D.bin is represented and how other 3DGS methods consume it. Done means EDGS can export a points3D.bin file that another 3DGS pipeline can use instead of the in-memory initialization.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, pytorch
Domain
computer-vision, machine-learning
Issue type
Feature
Difficulty
4/5
Estimated time
3-5 days
Activity status
Quiet
Clarity
Mostly clear
Newbie friendliness
48/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.