CC BY 4.0

Reading Room (No Lidar， Only Photos)

6.2K

610

No Lidar, Only photos were captured with Insta Ace2Pro. Gaussian was trained by MipMap Desktop. The performance from pure photos alone in a weak-feature indoor environment is genuinely surprising — no artifacts, no floating fuzz or "cotton-like" noise. That's actually one of the major challenges in Gaussian Splatting. I used an Insta360 Ace Pro 2 action camera, so the overall cost was kept pretty reasonable. MipMap Desktop: na.mipmap3d.com

Software Used

MipMap Desktop

Suggested

2 Comments

jex3 months ago

could you share the photos you used to train

will4 months ago

Nice scan! But this scene is scaled too large currently. In SuperSplat Editor, scale your scene so that 1 unit equals 1m. Then, Walk Mode will feel as though the head is at the right height.

Loading comments...