I am a new Mini 2 owner struggling to get comfortable with the Mini 2 and RevoScan 5.0. The included bust was super easy to scan, but I have so far not found anything that is even remotely as simple
Most of my needs are for reverse engineering and the parts are typically low on features and always have some kind of symmetry. So far I would have been much better of just drawing the parts in Fusion 360 without scanning first.
I accept these are not the easiest cases so I am trying all the tricks I have read on this forum, using the marker mat, adding bits of pieces to the parts to break symmetry, putting marker stickers on (disaster on small parts without flat surfaces), spraying with AESUB, turning off the room lights.
I have ordered the tilting table to hopefully make re-orienting a bit easier. I also hope the tilting table is synchronized with the scanner so motion smear is less of an issue.
Thoughts on RevoScan:
Every time I select a file the menu selection reverts to one click processing. It would be much more convenient if it just staid at the last selection.
In many of the processes it would be really nice (time saving!) if it was possible to select multiple files to be processed.
Every time I go into fusion the point distance changes to some new value, why is this? It would be more convenient if a previously selected value was kept for a project. I dont understand this process very well I suspect. From what I read, and the lastest instruction video, for maximum control and accuracy during manual processing one should choose standard and a suitably low point distance. In my case < 0.1
The feature merge is pretty bad at aligning, I have started playing with CloudCompare and its alignment is massively better. It should be possible to have a way to do a coarse alignment by rotating the scans and placing them on top of each other, and then press a button for the fine alignment in RevoScan.
The marker merge is a bit of a mystery as well, I try to select matching points, but in a point cloud that is generally pretty hard without some computational help to do the fine alignment. So far I have had little luck with this method.
Pretty sure I cannot see any difference on the point cloud isolation function when choosing different percentages. Does this work as it should?
There should be way to do bulk export to 3rd party tools such as CloudCompare
Please add SpaceMouse support - feels like I am missing a limb here
Any advice is welcome, I still have high hopes for one day being able to do a good hi-res scan with the Mini 2
For the alignment there’s no good features. Maybe you can put plasticine or something on it for better features? @PUTV is a specialist for such thing’s, she may can give some tips and tricks for that.
I am playing a little with CloudCompare and its clearly a better choice, however it also has quite a bit more of a learning curve
One thing I am curious about though is the minimum size of an alignment feature in RevoScan (if it uses “features” per se). It would be nice to know how the feature alignment works so that one could create optimal conditions, beyond “add some putty”
There is definitely still some room for improvement and revopoint is working on it.
I hope the improvements will come in the near future, I’m lso already waiting for it! But in this price segment I won’t find a better Scanner/Software combo.
In this case you can’t align this scans because you are actually missing the sides , you need at least 20% overlapped and at least 3 matching points ( features ) that exist on both scans .
Cloud Compare will not align it better either since you still missing the side scan , remember you need 3 angles of a scan not 2 , what you can do is secure the object vertically , scan one rotation and use it to merge all the points from front and back together with the scans you already made .
The object was actually stood vertically and scanned as suggested, the two depicted scans have at least 20% overlap (more like 70%) as well as identifiable features (holes in this case). The second scan is of the object flipped to the other edge and the merge is mainly to get a scan with all four sides closed.
The alignment fails in RevoScan, but succeeds with both CloudCompare and Meshlab. I found that my workflow is almost certainly going to involve CloudCompare and/or Meshlab as it also allows me to prune/trim the aligned clouds before any final merge, keeping the scan details clean.
I have a question of what constitutes a “feature” though? Any edge, line, protrusion? What is minimum size (area, height)?
And this is a really dumb question - Does 20% overlap mean that any scans need to cover 20% of the larger scans surface? So if I am building a bigger scan from many small ones I will have to be careful of the ordering as adding a small scan to a big one will not work , but stitching small ones to each other and incrementally building multiple bigger to merge is fine?
I still like the RevoScan for cleaning the point cloud though, nice interface and good results (apart from overlap currently being broken?). Being a relative newbie to scanning (I have tried some photogrammetry) I find the revopoint software a promising start that just needs a little refinement to be quite enjoyable. The opensource options are all far too much for most use cases.
Feature is automatic recognizing of unique feature on the merged objects , it require to have at least the same 3 features on each object , it also don’t like much symmetry as it easy confuse the algorithms.
See it like a face recognition.
There are not specific rules regarding the size, but the distance depending of the scanner FOV size , how smaller the FOV how closer the features need to be , in MINI situation around 3 cm apart at max .
The scanner projects a pattern on the scanned surface and any distortions created using the features are used as tracking point .
Cloud Compare is superior in this case and I do use it in most cases as well as it offers more than just simple alignment.
You also asked early about marker points .
When you use marker point mode and markers (stickers) the algorithms ignores the object’s features and symmetry , and it tracking only the markers as features, you need minimal 3 markers to be visible at the same time while scanning .
You don’t need to stick the markers on your object , you can simply stick it on any dark surface like for example black plastic turntable or simple black garbage back , put the object on top and scan it .
Regarding cleaning , it is recommended to clean overlapped points and lose points .
Check the Tutorial section in the forum , there is lots of great stuff to learn from .
@grukx As of today, RevoScan does not have a marker merge function. There is one that they call marker merge but really it’s a point to point merging mode. I already made a feature request in this post Perhaps it would be worthwhile for you to mention this under my post, so that the revopoint team can see that many people are missing that feature.
For a real marker merge abilities the software and hardware firmware need to be rewritten what really will not happen.
The terminology is the way it is , changing it now will only create confusion since lots of tutorials are referring to marker mode …
I know that the dev team is working on the option that allows you to place your own marker on the scanned surface and use it as reference , for that reason it is called market mode as this function was once available.
Actually to be clear the Feature mode uses point to point algorithms, semi automatic function .
Hopefully "marker " selection will be available soon as I am waiting for this for so long already myself …
