Quality & Triage

Quality & Triage

SplatKing assesses each capture as you go, so you can catch weak data before you spend time reconstructing it.

The session summary

When you stop a capture, a summary appears — it reflects how the session went and gives you the actions to share or open the result. For LiDAR, it also reports the COLMAP export outcome.

Per-frame quality

Each captured frame carries a quality assessment — a score and a band, drawn from motion and blur analysis at capture time. Tap a band to see what it means.

GoodSharp, well-exposed, low motion.Keep — ideal reconstruction input.
Read bands in contextQuality is relative to conditions. A capture in a genuinely difficult environment (low light, fast motion) will show more weak frames even when you captured well — read the bands in context, not as a pass/fail grade.

LiDAR: the COLMAP export outcome

For LiDAR sessions, the summary reports what happened with the on-device COLMAP export:

COLMAP export: written (N images, M points)Success. The COLMAP_Text_Model/ is ready to use.
COLMAP export: skipped — fewer than 3 tracked framesNot enough cleanly-tracked frames. Move more / improve lighting and re-capture.
COLMAP export: skipped — verification failed / system errorA write or integrity problem. Re-capture; if it persists, report it.

The export is fail-closed: you either get a complete, verified model, or none — never a partial one.

Triage workflow

1Glance at the session summary right after capture — re-shoot then if it looks weak, while you’re still on-site.
2Before reconstruction, filter out Poor frames using the per-frame metadata.
3For LiDAR, confirm the summary reported a written COLMAP model.

Next steps

Gaussian Splatting News on YouTube

The week in radiance fields, every Monday.

Gaussian Splatting News on YouTube

The week in radiance fields, every Monday.

Gaussian Splatting News on YouTube

The week in radiance fields, every Monday.

Gaussian Splatting News on YouTube

The week in radiance fields, every Monday.