Large Scan Complete, An Angelic Journey

Over the last 4 days I’ve been doing a lot of troubleshooting with the software Revoscan in order to capture this scan of a larger object. The angel statue stands 76 cm tall, and it’s wing span is
38 cm at the widest point near the top.

I struggled with this immensely, but it was largely my lack of experience - it was a real learning journey, but this is exactly why I challenged myself with it. For smaller scans that will easily fit on the smaller turntable, this device was a breeze. But I wanted to see what it could do. Let’s get into it.

The memory limitations of using the wifi option through the phone was where I was mostly hung up for days. Ultimately, this was not doable. Revoscan would not allow to resume scanning past a certain point, and this exceeded 3500 frames. I attempted at multiple sizes per batch, 1000 frames, 350 frames, a few in between - it didn’t matter. Once it came time to go back and add a new scan, Revoscan would not repopulate the scan I was working on, it would crash, or freeze and then crash.

Because this model is much wider than it is deep, turning the sculpture in order to capture it ran into errors every time the side face of the wing would pass the laser - it lost tracking because it could not realize that it was now looking at something quite narrow where as just a small angle before it was looking at something quite broad.

I will say on that front - it is imperative that you select “Manual” on the depth camera in order to get successful scans of objects this variable. Certainly when handheld capturing. That’s mentioned on this forum but it must be stressed to any newcomers - this isn’t optional, to the people constantly losing tracking, this is the issue.

Ultimately I found success when I brought the scanner back into my office and used USB mode with the windows version of Revoscan - I attempted capturing scans by turning the model itself on a larger turntable, but again the difference in width and depth made this impossible. I managed with the extra length of the cord to capture the troublesome areas (inside the wings and the edges of the wings) handheld. The model you see here has been captured handheld - albeit plugged into a computer.

Admittedly, if I had a room dedicated to scans and had a pole set up that could easily vary the height of the scanner and the model could be turned remotely without my physical interference - that would have made this much simpler. That is not a limitation of the device or software. However, I still would have struggled due to how close to the object one has to be in order for it to track, when it came time to view the side of the wing, it simply could not figure out what it was seeing.

But the results, oh, the results when you’ve finally managed to work WITH the device? Absolutely beautiful, in my humble opinion.

You’ll notice here there is an artifact on the back of the angel which I did not notice during my many multiple scans of the surface - I understand ideally you would make a single pass of the object, but due to the variability in its geometry, I often had to go back to a familiar space the scanner could remember correctly before scanning off in either direction. I am not concerned about this artifact, which can very easily be removed in post production.

Now you may be looking at the angels left hand thinking it’s a poor capture, but in reality this is actually incredibly accurate - this is precisely what the plaster model looks like, with the infill and the bulbous nature of the wrist. This I also will be able to sculpt easily in post production.

Finally, I’d like to say that when it came time to fuse and mesh this, my computer having the AMD Ryzen 5 4600G 3.70 GHz processor and only running 12 GB of RAM - it took approximately an hour and a half to process this model.

I suspect that is due to it being a little lower on RAM, but also because of how many duplicate scans I was taking of areas over and over again in order to complete.

Overall, this was a learning journey for me. I’m much more aware of the units capabilities, and the software limitations at present.

However, I am duly impressed with the quality of the model, and very much look forward to all my future projects.

EDIT: At the time of this scan, Revoscan was the version.


Thanks for sharing your experience

I am going to build the “DIY Slider” as posted in this forum for full body scans. It uses CNCjs for control of the elevation of the POP and rotation of the turntable.

1 Like

Impressive results. I look forward to seeing more of your work.

Wonderful result indeed ! well done @Swiftwater11

Impressive work!
How did you manage to get the entire model? I can not achieve a good work flow when it come to scan a large object that need - for me- multiple scan…
did you scan multiple part, and then assemble them in meshlab?

I attempted that method with some mixed results using the phone wifi handheld method collecting several scans and then aligning them in meshlab - however the revoscan software simply wouldn’t allow me to resume scanning with my phone.

I tried for about 4 days with that method and got a model that was about 70% done. But it left me with some post production nightmares I didn’t want to deal with.

Ultimately I was able to complete this by using the USB method plugging into my computer - the revoscan software version for windows DID allow me to resume scanning.

The only problem was the amount of faces and vertices of the final model, because I had to resume scanning from a position I’d already scanned multiple times - a very redundant model.

Fortunately Revo studio allowed me to simplify this model several times and made it much more manageable without losing any of the quality - it was able to remove the redundancy without compromising quality.

As i have not yet the SN for revo studio, I tried the method with meshlab. There is a function to align mesh and other to merge and simplify the model. and the you can export the result.
using one single scan for a large piecde like this looks tricky, with a lot a undo. But I discover than even with point cloud errors (alignment and some points not “on the model” ,
merging of the point cloud then meshing correct a lot of errors!
I hope revo studio is efficient!