how to obtain the "correct" depth maps from the output? Are the generated depth maps subjected to any form of "warp"?
Opened this issue · 6 comments
I've trained the model on my own dataset and achieved promising results. However, when I convert the RGB images along with the depth maps into point clouds, I noticed some spatial distortions. I suspect this might be related to the warp process.
@BaoZehan I'm having a similar issue. Have you solved this issue ?
There are plenty of issues related to this important factor for those who are interested to return point cloud out of rendered depth but with no answer as for now, for example, #80 and #101. I wonder if the returned points have been unwarped back to Euclidean space. If not, it would be very helpful if authors can refer us to the idea.
There are plenty of issues related to this important factor for those who are interested to return point cloud out of rendered depth but with no answer as for now, for example, #80 and #101. I wonder if the returned points have been unwarped back to Euclidean space. If not, it would be very helpful if authors can refer us to the idea.
#101 shows possible reasons for the problem