3DGS data preparation
Use the online COLMAP editor to inspect sparse reconstruction data, remove obvious outliers, review camera poses, and export cleaned points and images files. It is designed for the preparation step between image alignment and 3D Gaussian Splatting training.
Open COLMAP text or binary files directly in the browser and inspect the sparse point cloud visually.
Use rectangle and lasso selection to remove outlier points that can harm reconstruction quality.
Review camera poses, orphaned image entries, and point statistics before exporting cleaned files.
Removing outliers before training a Gaussian Splatting scene.
Checking whether COLMAP alignment produced coherent camera poses.
Preparing cleaner SfM data for NeRF, 3DGS, or photogrammetry experiments.
Debugging sparse reconstruction issues before spending time on full model training.
The editor supports COLMAP points3D.txt, points3D.bin, images.txt, and images.bin files. You can also add image folders so camera entries are easier to inspect visually.
Noisy sparse points, orphaned images, and bad camera entries can make training less stable. Cleaning the reconstruction before training helps reduce artifacts and wasted optimization time.
Yes. After selecting and deleting outlier points or problem images, you can export cleaned COLMAP text or binary data for the next stage of a NeRF or 3DGS workflow.
No. The browser reads your files into the editor. Changes are applied inside the workspace, and your original files remain unchanged until you export a cleaned copy.
