360 Utility

360Extractor

360Extractor is an open-source tool for extracting rectilinear perspective images from 360-degree video for photogrammetry and 3DGS datasets.

Stage: Utility360-to-COLMAP prepIndoor capturesOpen-source workflows

Open official linkGitHub Repository->

What It Does

360Extractor helps bridge 360 cameras and classic reconstruction tools. By turning a 360 video into many rectilinear views, it makes the capture easier for COLMAP-style pipelines to understand.

It is especially useful for room-scale spaces where a 360 camera captures enough coverage but the downstream trainer expects normal image geometry.

How To Use It In 3DGS

Capture stable 360 footage with minimal motion blur.
Use 360Extractor to generate perspective views at useful directions and intervals.
Run COLMAP or another alignment tool on the extracted image set.
Train the splat and use an editor to remove tripod, seam, or ceiling artifacts.

Things To Watch

Too many extracted views can create a huge and redundant dataset.
Poor 360 stitching or compression artifacts carry into the extracted images.
Extraction settings affect field of view and overlap, so test on a small sample first.

Official And Related Links

GitHub Repository COLMAP Documentation FFmpeg

Extraction Strategy

Treat extraction as camera design: choose virtual lens FOV, yaw/pitch directions, and temporal spacing intentionally.
Start with fewer directions and frames, run a quick COLMAP alignment, then increase density only where coverage is missing.
Keep the original 360 video; if alignment fails, you can regenerate perspective images with different settings.