Tips for people using RC, running into trouble and not finding enough documentation about it to figure out what to do.
I will be adding more to this as I encounter issues and overcome them.
If you have any more troubleshooting suggestions, feel free to reach out via Twitter DM.
Alignment immediately fails. “Feature detection completed in 0 seconds”
Solution: Reset Alignment settings and try again.
After resetting, you can change certain settings and try again. This time, it most likely will not immediately fail.
EDIT: A member of the RC team responded with a solution as well:
If you run alignment and there are no components present in the project, most likely you have “Merge components only” set to Yes in the alignment settings. This means that no new images are taken into account during alignment and the application merges only existing components. Therefore, if there are none present, RC cannot merge anything. In this case it is sufficient to change only this setting.
Reconstruction crashed overnight and restarted the PC. No crash log to know what happened exactly.
Solution: This most likely is related to the specs of your PC (too little RAM).
- One solution is to increase the reconstruction downscale factor (2, 3, or even 4), but you may lose certain fine grained detail.
- Another solution is to add a “minimal distance between two vertices”, thus setting a constraint on how fine grained the meshing will be. If your scan is set to 1:1 scale (if images have GPS coordinates in their EXIF, like from a drone or smartphone), then 0.002 is 2mm. RC’s units are always in meters.
Massive chunks missing from the mesh reconstruction (not because there were no cameras facing that region)
The problem arises when the dataset includes scanning an object/region very densely while combining it with drone photos of the surrounding region at a lower density.
In this case, it’s a densely scanned tree combined with a scan of the valley that it’s in:
- Either increase reconstruction “downscale factor”, increase “minimal distance between two vertices” (as seen in the previous problem)
- reconstruct the dense object and its surrounding separately using a reconstruction region.
Once you make a reconstruction region around the region, you can export it so that when you reconstruct the surrounding, you can reimport the reconstruction region, and select the triangles within the box to delete them, and combine the two meshes in whatever 3D software (Blender) or 3D engine (Unity). Having two separate meshes also helps you decimate and unwrap with different settings (so the object can have higher poly/texture than the surrounding)
More to come as I encounter weird problems!
Feel free to copy these solutions to whatever forum/thread that’s necessary. We need better SEO and documentation for unusual Photogrammetry problems!