When I transform my own depth map to .dmap file, DensifyPointCloud went wrong

[MarcoG5]

First I would like to thank you for your great work! But I ran into some trouble when I try to use my own depth map. I only have camera param, original image and depth map information and I don't have normal map or confidence map. I seems that the DensifyPointCloud can not generate normal map and confidence map properly. This is the normal and confidence map that I didn't use my own dmap file. After I use my own depth information, the normal and confidence map it generates become weird Did I do something wrong? Looking forward to your reply !

[cdcseacave]

pls send a small scene with only 2 cameras and dmaps, and the command line you use

[MarcoG5]

sparse.zip

[MarcoG5]

pls send a small scene with only 2 cameras and dmaps, and the command line you use

I simply used the command DensifyPointCloud scene.mvs

[cdcseacave]

The camera poses are not correct:

Pls use InterfaceCOLMAP for exporting from COLMAP:

[cdcseacave]

And here is the fused point cloud:

[MarcoG5]

Did you mean that the camera pose in my .dmap file is wrong? I did use InterfaceCOLMAP before I use DensifyPointCloud

[MarcoG5]

And here is the fused point cloud: I actually get this result by use InterfaceCOLMAP and DensifyPointCloud. I am trying to use my own depth map by writting .dmap format file. But using my own .dmap file I couldn't get this result. It confused me.

[cdcseacave]

the scene.mvs you sent me is wrong the dmaps you sent me have correct camera pose, but if you ask me the intrinsics are not correct

[MarcoG5]

the scene.mvs you sent me is wrong the dmaps you sent me have correct camera pose, but if you ask me the intrinsics are not correct

The scene.mvs I got it by using InterfaceCOLMAP. Did you mean that you used my .dmap file and get this fused point cloud？I will double check if any steps gets wrong and reach out to you soon. Thanks again for your help！

[MarcoG5]

@cdcseacave I still cant figure this out. This is a new example. I use COLMAP to get the camera pose and InterfaceCOLMAP command to get the scene.mvs It seems that when I use DensifyPointCloud without put my own .dmap file, the fused point cloud is correct. But if I use my own .dmap file (without normal and confidence map), the result went wrong. I use both DensifyPointCloud scene.mvs at the same time so I thought the problem is in my .dmap file? sparse.zip

[Bill-WangJiLong]

@cdcseacave 我还是想不通。这是一个新的例子。 我使用 COLMAP 获取相机姿势和 InterfaceCOLMAP 命令获取 scene.mvs 似乎当我使用 DensifyPointCloud 而不放置我自己的 .dmap 文件时，融合点云是正确的。 但是如果我用自己的.dmap文件（没有法线图和置信度图），结果就出错了。 我同时使用了 DensifyPointCloud scene.mvs，所以我认为问题出在我的 .dmap 文件中？ 稀疏.zip

Have you solved this problem？

[834810269]

我同时使用 DensifyPointCloud scene.mvs，所以我认为问题出在我的 .dmap 文件中？

@cdcseacave I still cant figure this out. This is a new example. I use COLMAP to get the camera pose and InterfaceCOLMAP command to get the scene.mvs It seems that when I use DensifyPointCloud without put my own .dmap file, the fused point cloud is correct. But if I use my own .dmap file (without normal and confidence map), the result went wrong. I use both DensifyPointCloud scene.mvs at the same time so I thought the problem is in my .dmap file? sparse.zip

Hello @thomas-19 ,

I have the same idea of utilising the measured depth map from an iphone or other depth sensors. It is much appreciated if you could answer my question.

@cdcseacave I still cant figure this out. This is a new example. I use COLMAP to get the camera pose and InterfaceCOLMAP command to get the scene.mvs It seems that when I use DensifyPointCloud without put my own .dmap file, the fused point cloud is correct. But if I use my own .dmap file (without normal and confidence map), the result went wrong. I use both DensifyPointCloud scene.mvs at the same time so I thought the problem is in my .dmap file? sparse.zip

Hello！How do you convert the depth map to .dmap?

Thanks in advance!