I'm going to be getting a 3D scanner

[Peter Grass]

Since I am not great at 3D modelling, I still wanted to give back to the community, I decided to order a 3D scanner so some of the older parts can be scanned since there is quite a lot missing from 2000 - 2020 from when there were no 3D models available. Some of the very old stuff was also hand measured so I dunno if those will want replacing since it depends how often anyone even uses those old parts really.

The scanner I will be getting is the Creality Raptor Pro.

So is there any particular preference of things that people would like scanned to start with?

If I scan the parts would someone be willing to get them ready so they can be added to the tracker?

Reducing the meshes to some kind of sensible size should be something I can do.

[Rene Rechthaler]

some of the bionicle stuff is quite freeform and not digitalized, that would help...
also, bionicle is quite cheap and has a lot of different parts.
(cant say anything about the scanner and 3d scanning, i only know that my earpieces for my hearing aids are scanned for manufacturing)

[Philippe Hurbain]

The scanner I will be getting is the Creality Raptor Pro.

Wow... pricey but looks pretty precise! Probably also easier to use than this   https://philohome.com/scan3dlaser/scan3dlaser.htm

[Peter Grass]

Wow... pricey but looks pretty precise! Probably also easier to use than this   https://philohome.com/scan3dlaser/scan3dlaser.htm

Heh, yeah not cheap by any stretch of the imagination but at least that one is made from actual lego.

Not really sure that amount of precision is needed but I guess its better to have too much than too little.

At least they have come down in price, a few years ago it would have been £10,000 if not more.

[Peter Blomberg]

The scanner I will be getting is the Creality Raptor Pro.

So is there any particular preference of things that people would like scanned to start with?

If I scan the parts would someone be willing to get them ready so they can be added to the tracker?

Reducing the meshes to some kind of sensible size should be something I can do.

Great! That is an amazing scanner.

My greatest desire is the Duplo candle https://rebrickable.com/parts/11854/dupl...ame-print/
It has been around for 15 years, but apparently not in BI.

I'm definitively interested in preparing any Duplo pieces for the tracker.

[Orion Pobursky]

I've been eyeing the OpenScan kit.

[Rene Rechthaler]

I've been eyeing the OpenScan kit.

would be much cheaper (especially if you already own a printer)

[Hageta]

I've been eyeing the OpenScan kit.

I actually have a OpenScan Midi, but I am not quite proficient with it yet.

[Hageta]

Building the Midi version was not easy, as the documentation was lacking in some places.
I think this is not the case for the other versions/sizes.

[Peter Grass]

Here is a quick test I did on the duplo candle and to see how easy it is to reduce the mesh complexity.  Obviously there are some holes I didn't fill in and I haven't scanned the bottom of the part or merged the scans. It was just to see how easy it is.

I didn't use any scanning spray but it's done a reasonable job of capturing the model even though its super shiney.



This is with 11K triangles which is still too many but better than 1 million previously.

[Philippe Hurbain]

Nice!!!

I didn't use any scanning spray but it's done a reasonable job of capturing the model even though its super shiney.

I had so-so results with scanning spray (improves reflections, but tends to accumulate in concave corners). That was a while ago, maybe the spray improved since! The method that works for me is to dip the part into baby powder, then blow it out completely. Only a very thin dust veil remains that does the job for shininess (or transparency) problems.

[Peter Grass]

Nice!!!
I had so-so results with scanning spray (improves reflections, but tends to accumulate in concave corners). That was a while ago, maybe the spray improved since! The method that works for me is to dip the part into baby powder, then blow it out completely. Only a very thin dust veil remains that does the job for shininess (or transparency) problems.

Thanks I will give that a go when I scan it properly, I also need to get some blutak or something to hold the parts in place for scanning the bottom side.

[Peter Blomberg]

Wow! Very nice indeed!

Did you measure it standing up or lying down? Is the "other side" of the candle as good at the top?

The inside is relatively easy to model without any scan. It's the top that is challenging. The part isn't symmetric.

A mesh size with vertexes 4 LDU apart would be sufficient, 2 LDU if you want to create high res. Just for reference, a 48-sided cylinder would have the vertexes approximately 2.1 LDU apart at the outer rim (R16.6) of the candle.

[Peter Grass]

Wow! Very nice indeed!

Did you measure it standing up or lying down? Is the "other side" of the candle as good at the top?

The inside is relatively easy to model without any scan. It's the top that is challenging. The part isn't symmetric.

A mesh size with vertexes 4 LDU apart would be sufficient, 2 LDU if you want to create high res. Just for reference, a 48-sided cylinder would have the vertexes approximately 2.1 LDU apart at the outer rim (R16.6) of the candle.

I did it standing up but when i scan it properly I will do it on its side as well and a bottom scan and combine all 3.

Of the current scan both sides are roughly the same quality, I can go better though and try to fill in the gaps as much as possible even if it will get reduced down eventually.

The insides are basically hollow and you can see right to the pinhole on top when you look inside.

[Peter Blomberg]

Of the current scan both sides are roughly the same quality, I can go better though and try to fill in the gaps as much as possible even if it will get reduced down eventually.

You sure like to overdo it. Me too sometimes. Here, it might not be necessary. The existing scan is more than sufficient for LDraw. No need to fill in those tiny gaps. I recommend you try to reduce the mesh further without doing any more scans or filling up holes. Getting down one order of magnitude to under 2k triangles would do more than well enough.

[Peter Grass]

You sure like to overdo it. Me too sometimes. Here, it might not be necessary. The existing scan is more than sufficient for LDraw. No need to fill in those tiny gaps. I recommend you try to reduce the mesh further without doing any more scans or filling up holes. Getting down one order of magnitude to under 2k triangles would do more than well enough.

Well I am just kinda using it for practice since its the first thing I have scanned.

[Peter Blomberg]

Want more challenges?

Duplo map https://www.bricklink.com/v2/catalog/cat...P=54073px1
Not in BI and has helical curvature which is hard to measure by hand.

Duplo seaweed https://www.bricklink.com/v2/catalog/cat...ge?P=43852
BI mesh is inadequate. It has a wavy shape hard to measure by hand.

Duplo catapult...

[Peter Grass]

Here is the candle reduced to around 2000 triangles which you can use if you want. Well its 4000 lines excluding cond lines but I definitely selected 2000 when I was reducing it.

I'll order those duplo things when I order my next set of parts. I was going to get the simpsons kwik-e-mart heads that ldraw is missing and scan those.

[Gerald Lasser]

Here is the candle reduced to around 2000 triangles which you can use if you want. Well its 4000 lines excluding cond lines but I definitely selected 2000 when I was reducing it.

I'll order those duplo things when I order my next set of parts. I was going to get the simpsons kwik-e-mart heads that ldraw is missing and scan those.

THe candle is actually doubled in the file, when rotating LDPE doesn't show it, but panning down you see the second candle below

[Peter Grass]

THe candle is actually doubled in the file, when rotating LDPE doesn't show it, but panning down you see the second candle below

Ah oops I think I used the wrong option when moving it to the origin when I was checking it out in blender. Thanks!

I removed the duplicate now and trimmed the bottom edge a bit.

[Peter Grass]

I scanned the koala part 2589pb01 someone wanted here
https://forums.ldraw.org/thread-29079.html

In case anyone wants to do anything with it.



This one is a combination of 3 different scans at different positions as its much harder to scan small parts and ones which have shadows due to parts of them being close to other parts (like the hands)

I think it turned out OK. 3300ish polygons since its a organic shape so needs to be somewhat curvey.

[Philippe Hurbain]

Pretty good results!
The good thing is that the koala is also available in the Building Instructions so we can compare. Perfect size, no significant distortion. As expected, the biggest problems occur in the sharp edges that are somewhat blunt by the molding process and by the scan.
Here is the BIs mesh and a comparison file where I manually aligned your mesh.

[Peter Grass]

Pretty good results!
The good thing is that the koala is also available in the Building Instructions so we can compare. Perfect size, no significant distortion. As expected, the biggest problems occur in the sharp edges that are somewhat blunt by the molding process and by the scan.
Here is the BIs mesh and a comparison file where I manually aligned your mesh.

I think some of that may be down to the smoothing settings I was using in the app I guess it will need some tweaking to see what works best. I didn't find it in BI when I looked which is why I scanned it but its nice to get some conformation it looks decent.

[Peter Blomberg]

The Koala is a small part. Only about 16 mm high. If one wants to see the sharpness of the rounded corners, one needs sub-LDU resolution, which is about 3-4 times as many triangles as you have now. Most of them will not be needed in the end, but it seems mesh reduction isn't as selective as one would wish.

Can you try the color scan on the koala?

[Peter Grass]

The Koala is a small part. Only about 16 mm high. If one wants to see the sharpness of the rounded corners, one needs sub-LDU resolution, which is about 3-4 times as many triangles as you have now. Most of them will not be needed in the end, but it seems mesh reduction isn't as selective as one would wish.

Can you try the color scan on the koala?

I tried the colour scan on both the koala and the candle but they are too shiney for the colour to be picked up. If I was scanning something like stone instead of plastic it would probably work. When scanning colour you are also using the lasers which shine on the object which makes them even more shiney so its a catch 22 situation. If you coated them in scan spray that would also hide the colours so that wouldn't work either.

There are apps to selectively decrease the amount of polygons while keeping the detail but they are crazy expensive since they are meant for game development so certainly not in my budget. You could maybe manually do it in blender but that would take ages and not be a fun task.

[Peter Blomberg]

There are apps to selectively decrease the amount of polygons while keeping the detail but they are crazy expensive since they are meant for game development so certainly not in my budget. You could maybe manually do it in blender but that would take ages and not be a fun task.

We certainly can write some script that does a decent job at mesh reduction. The trick is to remove the vertices where the normals of the surrounding triangles are as parallel as possible. This is a lot more calculation-intensive that just averaging the vertexes of selected non-touching triangles.

For the algorithm:

1. For every triangle, calculate the surface normal.
   
2. For each vertex, calculate the sum of all angles between adjacent normal pairs around that vertex. The number of triangles doesn't matter, as doesn't the curvature "through" the vertex. The sum will catch all that and more.
   
3. For each triangle, calculate the sum of the sum_of_angles of its three vertices.
   
4. Sort the triangles and their sum_of_sum_of_angles.
   
5. Determine a suitable cutoff or a suitable number of triangles to remove.
   
6. Begin looping thru each triangle, starting from the smallest sum_of_sum_of_angles. Check that none of its three vertexes are embargoed. Remove the triangle by averaging its three vertices. Adjacent triangles where two of its vertexes merge are removed at the same time. Place all surrounding vertexes in an embargo, i.e. they cannot be removed during this round.
   
7. At some point, looping thru triangles will no longer yield anything reasonable to remove. However, the cutoff should have been encountered long before such happens.
   
8. Go back to step 1 and repeat. Updated normals need to be calculated only for triangles that have been modified by the merging of vertices. All embargoed vertices need updated sum_of_angles.
   

Some additional notes: In the general case, it is impossible to remove just one vertex from a mesh because one cannot guarantee that the surrounding vertices are in a convex shape. Furthermore, any new triangles might become very long and narrow, which is highly undesirable.

[Peter Blomberg]

Thanks Peter, the candle is now on the tracker.
https://library.ldraw.org/parts/49747
https://library.ldraw.org/parts/49748

The entire cylinder with collar and interior has been made out of primitives. The scanned mesh didn't really affect those. It was quite tricky to turn the part exactly vertical. Partly because the cylinder in the mesh wasn't perfectly round. Furthermore, LDPE uses the y axis instead of the z axis for vertical.

The flame was almost completely guided by the scanned mesh. Unfortunately for me, there were a few downsides to reducing the amount of triangles to such a low number. The hole on top was smaller in the mesh due to the lack of scanning depth inside the hole, which allowed the outer surface vertexes to dominate the edge points during the rounding/reducing algorithm. Secondly, the tiny geometrical features (size about 1 LDU) at the interface between the yellow and white parts above the outer collar were smoothed away. These had to be recreated manually. There were also some rounding to the curvature of the flame near the ledge, which had to be manually corrected.

One sixth of the flame had some wavy vertical stripes that needed manual correction/smoothing. The number of triangles that needed to be flipped was quite low and the number of vertexes that had to be merged was also low.

I can easily see a workable pipeline from a scanned mesh into a geometry ready for the tracker. My only wish for the future would be to retain more triangles before switching over to LDPE since that would enable the accurate modelling of also smaller features. The downside of this may be the uneven reduction of triangles needed on the scanning side. If need be, I could work with two files of differing number of triangles to get the best out of both.

The edge lines and cond lines in the file were effectively useless. I deleted them very early on and recreated only the cond lines I needed in the end using Edger2 in LDPE.

[Peter Grass]

Nice work on the candle!

I will upload higher polygon meshes from the next things I scan so then people can get more detail since otherwise it isn't worth scanning parts if its going to end up worse quality than the BI meshes when it gets reduced too much. Probably 12K polygons or maybe a bit more for parts that have lots of curves / organic shapes. Then people can pick and choose what they keep / reduce.

I got the simpsons heads and also a few parts that people were requesting and I will scan them when I get chance. Those Duplo parts are going to end up being expensive except the pond weed as the shipping costs are insane even when they are coming from Europe (they aren't available from sellers where I live in the UK)

I was also thinking of scanning something larger like 44510pb03 which is one of the raised baseplates which ldraw doesn't have. Something like that would be a pain to make manually and a scanner is built for doing stuff like that and the meshes wouldn't need to be too high in the end result since apart from the rocky terrain the slopes wouldn't need many polygons in the final result. Plus since those are vacume formed plastic I want to see if that makes any difference to the colour capture of the scanner. Since its also larger it should be easier to scan it from further away meaning less shine on the part I would have thought.