cdcseacave / openMVS

open Multi-View Stereo reconstruction library
http://cdcseacave.github.io
GNU Affero General Public License v3.0
3.39k stars 911 forks source link

Densifypointcloud generates very less points. #1125

Closed kundu03 closed 7 months ago

kundu03 commented 8 months ago

Hi @pmoulon and @cdcseacave,

I am using DensifyPointCloud to generate a dense point cloud from a sparse point cloud generated using OpenMVG.

The quality of sparse point cloud generated from openMVG is really great with a total feature/points count of 203,400.

But, when i try to densify it using DensifyPointCloud, i am getting only 43,696 points. Can you please tell, what could be the reason for this and how to solve it. Thanks in advance.

Here is the log of DensifyPointCloud usage.

12:49:41 [App ] OpenMVS x64 v2.3.0 12:49:41 [App ] Build date: Mar 16 2024, 10:17:25 12:49:41 [App ] CPU: 12th Gen Intel(R) Core(TM) i9-12900K (24 cores) 12:49:41 [App ] RAM: 62.57GB Physical Memory 2.00GB Virtual Memory 12:49:41 [App ] OS: Linux 5.15.0-101-generic (x86_64) 12:49:41 [App ] Disk: 412.37GB (959.67GB) space 12:49:41 [App ] SSE & AVX compatible CPU & OS detected 12:49:41 [App ] Command line: DensifyPointCloud /home/Desktop/openmvg/openmvg2openmvs/Test6/openmvs/scene.mvs -w /home/Desktop/openmvg/openmvg2openmvs/Test6/openmvs/ --dense-config-file /home/Desktop/openmvg/openmvg2openmvs/Test6/openmvs/Densify.ini --resolution-level 0 --number-views-fuse 2 12:49:41 [App ] MapSMtoCores for SM 8.6 is undefined; default to use 64 cores/SM 12:49:41 [App ] CUDA device 0 initialized: NVIDIA GeForce RTX 3050 (compute capability 8.6; memory 7.79GB) 12:49:42 [App ] Camera model loaded: platform 0; camera 0; f 0.483x0.483; poses 53 12:49:42 [App ] Scene loaded from interface format (103ms): 53 images (53 calibrated) with a total of 1175.26 MPixels (22.17 MPixels/image) 203400 points, 0 vertices, 0 faces 12:49:42 [App ] The camera directions mean is unbalanced; the scene will be considered unbounded (no ROI) 12:49:42 [App ] Point-cloud composed of 203400 points with:

cdcseacave commented 7 months ago

pls share the full data to be able to debug

kundu03 commented 7 months ago

Hi @cdcseacave ,

Here is the drive link to sparse point cloud file along with undistorted images and scene.mvs file.

https://drive.google.com/drive/folders/1EnWlD8YoQAQ7svBgqasl_5v2yx0wo3KR?usp=sharing

Thanks

5VK8Y commented 7 months ago

OpenMVG output is misaligned.

cdcseacave commented 7 months ago

indeed, I confirm, the images do not view the sparse point cloud in the right place, something happened during OpenMVG reconstruction or export

cdcseacave commented 7 months ago

all works fine is the SfM is done and imported correctly. This script is doing everything for you:

python3 OpenMVS\src\scripts\python\MvgMvsPipeline.py images openmvg

image