Scale of the reconstruction compare to real world?

[imakjaiswal]

I'm seeking clarity on the function of the scaling factor within the Alicivison Meshroom structure-from-motion pipeline. I've reconstructed my data multiple times and obtained a point cloud, but I'm noticing a significant disparity in scale compared to the original LiDAR point cloud.

Could you clarify whether camera intrinsic properties, such as focal length and sensor width, influence the reconstruction process?

[natowi]

If you need your data scaled to real world scale, use Marker Scaling or GPS scaling.
https://github.com/alicevision/Meshroom/wiki/CCTAG-scaling
https://github.com/alicevision/Meshroom/wiki/Model-orientation

Each reconstruction run will be slightly different, especially if the initial pair is chosen automatically.

[imakjaiswal]

I've used the CCTag3 descriptor type, but unfortunately, the landmarks aren't being generated accurately as depicted in the screenshot attached. Could you please provide some suggestions for improving this situation? Additionally, I'm interested in understanding how CCTag3 or CCTag4 can be utilized to enhance the scale of the Structure from Motion (SfM) point cloud in comparison to the real world.

Please shed some light on this.

[natowi]

Ok, only 3 images...
https://github.com/alicevision/Meshroom/wiki/CCTAG-scaling scroll down to Improve CCTAG detection

[imakjaiswal]

[22:25:13.636169][error] Invalid regions in 'C:/Users/Trimble-WS04/Desktop/C&D Waste/S08-CCtag-2/MeshroomCache/CameraInit/04a9c17f095b0f33dfa49c0834510a54eaf14fe0/cameraInit.sfm'

Getting this error while running FeatureMatching2 node. Also how to avoid the geometric validation in FeatureMatching2 as the in attributes setting i am getting only these settings

Can you please help. Thankyou in advance

[natowi]

You can enable advanced parameters here: