princeton-computational-imaging / hndr Goto Github PK

Hi,
Why are normal maps in the paper so homogeneous (collapsed to a single color, corresponding to a planar and frontal depth map)?

The ones I get after backprojecting depth:

vs the ones in the paper:

I have had similar issues when depth is too small numerically and computing normals from depth/point-cloud. The issue in my case (which produced similar bad normal maps) was that when normalizing vectors before the cross product, I had an additive epsilon to avoid dividing over 0. This additive epsilon dominates if the depth values are small, so it requires adjusting the epsilon.

Hi! Thanks for the great work. I wonder what's the camera coordination convention(i.e. OpenCV / Colmap) used in this project?

World to Camera Transform: 
 [[ 0.13201442  0.02732324  0.99087113 -0.03884006]
 [ 0.401116    0.9126483  -0.07860719 -0.16735849]
 [-0.90646446  0.40783158  0.1095229   0.24308646]
 [ 0.          0.          0.          1.0000001 ]]