The point cloud model obtained by fusing depth maps generated by other methods is incomplete

[1z2z]

I generated depth maps by other means and wrote them to dmap format, to fuse them into point cloud use command line behavior: DensifyPointCloud -i scene.mvs -o scene_dense.mvs --geometric-iters 0 --postprocess-dmaps 0 --resolution-level 0 --max-resolution 3200 -v 4 But the fused scene is not complete, but I provided the depth map of other parts of the scene, what is the reason?Looking forward to your suggestions. (Supplementary Note: The scene.mvs file is converted from the colmap format. And I suspect it is a problem with the scene.mvs file, because the scene.mvs I converted using openmvg does not have this problem)

[1z2z]

Because of the file size limitation, I will send the scene and dmap files I use to your email, thank you very much!

[cdcseacave]

pls share them using google drive or some other online method, and not on email

[1z2z]

Hi, I have uploaded the file to google drive, the link is https://drive.google.com/file/d/1njZQNeKd3vlIdoCTiyC_FwK_rxBlULG2/view?usp=sharing Looking forward to your reply, thank you.

[1z2z]

I use Viewer to open the generated scene_dense.mvs and find that there is a frame outside the model. Is this the cause of the problem? how should I solve this?

[1z2z]

          Hi, I have uploaded the file to google drive, the link is https://drive.google.com/file/d/1njZQNeKd3vlIdoCTiyC_FwK_rxBlULG2/view?usp=sharing           I use Viewer to open the generated scene_dense.mvs and find that there is a frame outside the model. Is this the cause of the problem? how should I solve this?

          Looking forward to your reply, thank you.

^o^ @.***

 

------------------ 原始邮件 ------------------ 发件人: "cdcseacave/openMVS" @.>; 发送时间: 2023年6月6日(星期二) 凌晨0:36 @.>; @.**@.>; 主题: Re: [cdcseacave/openMVS] The point cloud model obtained by fusing depth maps generated by other methods is incomplete (Issue #1002)

pls share them using google drive or some other online method, and not on email

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

[1z2z]

I seem to have fixed this by setting estimate-roi 0

[cdcseacave]

the fusion is using only points visible in other depth-maps as well, with a param number-views-fuse set to 3 by default in order to filter out the outliers, if you trust the depth-maps you can decrease this param to get more completeness

[834810269]

I generated depth maps by other means and wrote them to dmap format, to fuse them into point cloud use command line behavior: DensifyPointCloud -i scene.mvs -o scene_dense.mvs --geometric-iters 0 --postprocess-dmaps 0 --resolution-level 0 --max-resolution 3200 -v 4 But the fused scene is not complete, but I provided the depth map of other parts of the scene, what is the reason?Looking forward to your suggestions. (Supplementary Note: The scene.mvs file is converted from the colmap format. And I suspect it is a problem with the scene.mvs file, because the scene.mvs I converted using openmvg does not have this problem)

May I know how you write a depth map into the DMap format? Could you please share the process?