radicaledward101

Sense Scanner

Sense 2 Scanner

A while back my girlfriend’s step-dad leant me 3D Systems’ Sense 2 scanner. I similarly got access to an old MonoPrice MP Select Mini 3D Printer from a friend who had upgraded quite a while ago. So I theoretically have both ends of the 3D replication pipeline now. Unfortunately, just like with the MP Mini, there are some limitations. And, while the MP Mini has (or at least had) a thriving community to support it with guides, tutorials, hacks, and mods, the Sense 2 does not seem to have the same support.

This lack of support is perhaps because the practical value of the Sense 2 is much lower than the MP Mini. But it is also likely that the anti-consumer practices of the parent company are to blame. The Sense 2’s hardware support ended on December 31, 2020. Software support will end on December 31, 2022 per the 3D Systems support page for the device. While maintaining official support for an old piece of hardware (this device was released in 2016 per this review) should not be expected, the software support coming to an end is much worse for one key reason: The software requires the original owner to provide an activation code via the account which purchased the device in order to connect and run a scan. This means, even at the time of release, the device effectively can not be resold and used as designed. And, unless a key-less version of the software is released, no one will be able to use the devices on a fresh install after 2022.

The Sense 2 uses an Intel RealSense Depth Camera under the hood. So the company that makes the device may have locked down the software simply because that was their real product. The camera itself was available in multiple other form factors, some direct from Intel.

This anti-consumer resale prevention would normally be enough for me to dismiss the product. But, since I have access to one for free, I wanted to put it through its paces and see what it could do anyway. Perhaps I will find another scanner that I’m interested in. Having this as a baseline for comparison will be very useful.

Resolution Issues

The biggest downside of this scanner is its supported resolution. It is accurate only up to 0.9mm on the x and y axis and 1mm in depth. This makes it pretty much useless for any kind of precision parts design work. It could be useful for some larger scale items for which a little bit of finishing work can be done to make up for any differences. But I thought I could use it for less precise work, like concept design, or, possibly most promising, miniature making.

My first scan attempt

My first successful scan was of my girlfriend at the head and shoulders level. The journey to run that scan was long and complicated, with a lot of dead ends. In short, while 3D Systems technically still provides software support, their software has some confusing bugs. It does not include everything in a single install. The driver and scanning software must be installed separately, and their documentation does not do a good job of explaining this. Once everything was installed, the device was not completely plug and play, due to an issue in Windows 10. I have provided simplified instructions at the bottom of this article in case you need them.

When I did finally get scanning working, I conscripted my girlfriend to act as a model for my first scan, in part so I could show off that I got it to work at all.

The results were underwhelming. Neither of us really liked how it came out. It was only a first attempt, but after doing so much work to set it up, and getting such a poor result, I developed an extreme mental block to returning to the project. This scan was actually performed in April of 2021, but I put off moving forward with it for almost a year.

So what exactly was wrong with the scan? First, the scanning itself took quite a while, during which she moved a bit and her expression changed. So that introduced some weird asymmetry. I also introduced some user error by failing to scan the top of her head and under her chin and ears. But the device itself is inaccurate enough that some more complex areas of her head artifacted pretty severely, particularly her ears, nose, and mouth.

Printing from the first scan

Now, after putting it off for almost a year, I’ve returned to the project. My main goal was to see if I could take something all the way through scanning to print. Since I already had the scan, I decided that the first step would be to see if I could take that and print it at all, before I went through the full process end to end.

Luckily the scanner saves the scan as a combination of obj for the model, a png for the textures, and mtl for applying the texture. Since I’m printing in a single color, I can just use the obj file.

To get an idea of what the process of taking a scan to print would be like, I dumped the obj file straight into Cura. I was surprised to see that some of the worst artifacting was actually only present in the texture, which gave me some hope of success.

Unfortunately, the obj is non-manifold. So Cura complained. I decided to follow the remesh + decimate solution provided by TLousky on the Blender StackExchange:

  1. Start new blender file
  2. Delete sample block
  3. Import obj file
  4. Select model
  5. Apply Remesh modifier
    • Switch mode to smooth
    • Set Octree Depth to 8
  6. Apply Decimate modifier (I’m not sure how important this actually is)
    • Choose Collapse
    • Lower ratio to decrease file size (TLousky set it to 0.05 but this seems low) mess with it until you find something that is a good balance of file size and rez.
  7. File -> Export -> Stl

The resulting stl file imported and sliced in cura just fine! So I decided to try printing it just for the hell of it. I made it small so it will print quick. Of course the mp mini rez means that this was not a clean print, like you would get from a resin printer or a very high end FDM printer, but it gives me a good idea of what the possible results will be.

I was pleasantly surprised by the result! I think, possibly because I scaled down the print so significantly, the detail of the scan was more than enough. I’m still not sure I would want to try to scan something and print it 1 to 1, and I definitely don’t think it would work for mechanical features, with the low resolution, and the remesh step in the pipeline, but this could definitely work for making some unique miniatures!

Attempt 2

With the first successful scan to print complete, I decided to scan a much more difficult object. For this I picked the Halo 3 Legendary Edition Helmet. It has much different geometry from a human being and also features one major reflective surface, which I had read causes issues for most scanners. At the same time it is of similar size and general shape to a human head. So further testing with mechanical and organic objects might be a good idea.

Halo 3 Legendary Helmet

My first impression was that, in the scanning tool, the scan looked surprisingly good with the textures in place. This was followed quickly by laughing at reflection issues, which while not unexpected, were still quite hilarious. I could see these issues forming during the scanning process, but seeing them in the final result was still surprising. The blobs that resulted as the scanner lost all concept of depth in the reflective surface are hilarious!

Scan with texture

Scan with texture 2

Scan with texture 3

Upon closer inspection I was suprised by how mushy the scan was. I pulled the obj into Blender for the remesh step as before. With the textures removed it became clear just how much lack of clarity there was in the crisp lines of the helmet.

Mesh unedited

Prior to running the remesh I decided to do a “quick” cleanup of some of the worst artifacting in Blender. I also cut off the chair that I had used as a stand for the scan. I am by no means a Blender expert. So this ended up taking much much longer than I had planned. The end result is more of a patch than a cleanup. I added essentially flat surfaces to the worst of the reflection issues and basically left it at that.

Mesh cleaned up

With the cleanup and remesh done, the import and slicing in Cura went just fine. I decided to print at around 15% size so that I could see the results in a decent amount of time. I’m also using some support settings that may not be the best, but I haven’t gotten around to finding supports that I like better yet.

Print with supports up

I definitely don’t like these support settings. So I’ll be changing those for future prints… if I remember.

Print with supports removed

Removing the supports was tough but I think the result was worth it. This was a surprisingly good result for my lowly FDM printer. The mushiness of the model is definitely still apparent but what detail is present does look good due to downscaling to 15% of the size of the original. Even my patches don’t look too bad! If I had a high quality resin printer I might be able to take some scans of lifesize people with this and make some really nifty 28mm miniatures.

At some point I may try painting up these two prints to see how they look. Unfortunately the supports have left some nasty seem lines and the mushiness is still going to impact any painting I do. I don’t yet have the skill to cover up a bad sculpt with good paint.

3D Scanning Moving Forward

For 3D scanning, at last check, there were still no really good solutions on the market for my purposes and budget. I would recommend waiting before buying anything in this space. Most of the scanners on the market either are toys that hook up to a smart phone or cost $600 or more. Even these produce scans with only slightly better resolution than the Sense 2, and with buggy proprietary software. Meanwhile, the high end scanners which are being used in museums and manufacturing are still sitting with $5k, $10k, or even $30k price tags.

Just as with the 3D printer industry, I expect that we will see many improvements and many price reductions, in the coming years that will change the landscape, and make 3D scanning a worthy addition to the individual maker’s toolkit. Right now, we simply aren’t there yet.

Appendix and Miscellanea

What follows are technical details on using the device and other notes that didn’t fit in the main article.

Install

I’m mostly including these instructions in case you already own one of these and want to give it a try. I would not recommend you purchase one at this time, especially given the impending end of life for software support. If you do need a copy of the software downloads, and can not find them archived elsewhere, contact me and I’ll see what I can do (note that there is not currently contact info on my site. I know. I’ll fix that someday maybe?).

  1. Install sense software “UnitySenseSetup-2.0.321.exe” from 3D Systems’ Website for Sense 2
  2. Install “intel_rs_dcm_sr300_3.4.98.4970.exe” from Intel’s Website
    • To do this I ran the command intel_rs_dcm_sr300_3.4.98.4970.exe --ignore-fw-update --silent --no-progress --acceptlicense=yes from powershell rather than using the gui, but you’re probably fine with using the gui.

Technical Issues

There is an ongoing issue with the Sense 2 on Windows 10 where its drivers will not pick it up properly and therefore even the official software can’t find it. Intel’s support site for RealSense has an article on the issue, but their solution did not work for me. In the end, the actual best way to fix this for me was to do the following:

  1. UNPLUG scanner
  2. ctrl-alt-delete
  3. Open Task Manager
  4. Go to Services Tab
  5. Find RealSenseDCMSR300 in list
  6. Right click -> Start
  7. PLUG scanner
  8. Start scanner app

I believe I found this solution on another blog or forum post, but I can no longer find the reference for it.

Scanning Software I Did Not Try

Theoretically any software that supports RealSense cameras should be able to connect with the Sense 2, but I was only able to find one free offering that did so, other than the software provided by the manufacturer. That was itSeez3D which is limited to scanning human beings.

I found two other paid options for scanning software listed here:

I would not recommend paying for software to run a Sense 2. You would be better off saving that money to purchase a newer and more widely supported scanner. After evaluating all the options, I decided to stick with the software that came with the device.

Other resources

Nick Lievendag did a much more in-depth review of the device. Nick also has more experience with scanners in general and used some techniques that I did not try. However, Nick is not focused on printing the resulting models. Of most interest here is the tracking pdf, the lighting equipment, and the turntable plus tripod setup.

Software Used