Matter & Form THREE - Second Example Scan (Shoe)

 Well before my interest in 3D, I was sometimes called upon by investors to help them determine the potential pros and cons of investing in a particular endeavor or product.  This process is called "Due Diligence" and the skill most needed is observation to details.  It is something I enjoyed doing and continue to apply those skills when examining new products entering the marketplace.

The goal isn't to find THE perfect product.  They do not exist.  The goal is to determine whether or not a product presents a good value for the investment to be put into it.  That includes taking into account the people behind the product.  Right now, my focus is on the Matter & Form THREE 3D scanner.  And, the team behind this product is one reason for my interest.  I have experience using not one; but, two Matter & Form scanners in multiple classroom situations.  Moreover, I have had excellent response to my questions regarding the best practices using those scanners.

The original Matter & Form scanner is not a perfect solution.  But, when compared to all of the scanners we owned at YouthQuest Foundation and Phillips Programs, including the Einscan, it was the easiest to use and provided the most educational value because of the scanning strategy utilized by Matter & Form.  The software not only encouraged experimentation; but, provided the feedback students need in order to continually improve their results.

Now we come to the place where this same company has announced a new product offering much higher resolution and much higher accuracy.   And, I have to believe that the solution and accuracy not only applied to the underlying mesh; but, with two 13mp Sony imaging chips, the material captures and wrapping as well.

But, for now, all we can do as pertains to performing our due diligence is to rely on the scan examples we have so far.  And, neither, so far, have included capturing the materials.  That's OK.  Because, it lets us focus on the underlying mesh with no distractions.  Here is a link to the original scan that Matter & Form uploaded to Sketchfab.

https://sketchfab.com/3d-models/three-hilfiger-shoe-mesh-b5275a878dfa4181b9dac44e3e436f56

I downloaded the .STL version from Sketchfab so that I could evaluate the scan using MeshMolder, a wonderful inexpensive application.  

Matter and Form also posted an image of the original shoe on the Reddit 3D Scanning forums.  Without the original image we have no way to ascertain just how accurately the THREE 3D scanner replicated the shoe.   I cropped the image to better zoom in on the details.

Original Scanned Show

Having the image of the original immediately cleared up one of my reservations at first glance.  It was not a smooth leather shoe; but, a cloth shoe, which accounts for some of the texture along the sides.  But, even so, there appears to be more texture than actually existed.  Fortunately, when I pointed this out on the Reddit forums relative to some artifacts in the coin scans, I immediately got this reply.

CubifyFan
6 days ago

I did notice some noise in places that are smooth in the coin. I've always been a bit bothered by smooth surfaces not being scanned as smoothly as they are in real life. Accurate capture of NON-features should be just as important as capturing detailed features.

I suspect this is capture device noise, which we also face when using high ISO in digital cameras. Experienced photographers sometimes use layering or stacking of multiple images to reduce noise by pixel averaging across the layers. I'm wondering if we could be given a choice to use a similar 'layered multi-scan' technique to cancel noise in a 3D scan to smooth flat areas without compromising actual detail.

drewshark
6 days ago

Yes it's noise from the cameras, you're right. We've identified three major sources of noise in our scans. And your idea of multi layering is something we want to explore to compensate for one of the sources. Thanks for the great idea!

That shows me that they are already on the issue. So, for now, I'm not too concerned about it.

So, let's look at some specific areas of the shoe to see what we can gleam from this example.

THE SOLE

A good place to start in evaluating the accuracy of the scan is the are around the front of the sole.

Image of the Shoe Sole

Scanned Result - Shoe Sole

When we compare images, it's plain to see that the THREE was able to capture the texture of the sole better than any previous scanner we'd used.  There is some roughness around the area of the black stripe; but, look at the capture of the end of the shoe string!  And, the transition ridge around the top where the fabric meets the rubber sole!  Those are excellent results.  

SHOE LACES

Image - Shoe Laces

THREE Scan - Shoe Laces

The THREE made a valiant effort to capture the texture of the shoe laces. But, the movement of the laces during the scanning process conspired against it.  Here are the notes from the Sketchfab upload.

Captured with MAF THREE Using a Turntable twice. Once with the sole down on the turntable, 360 degree, 9 captures Once with the sole facing out, 180 degree, 5 captures. Aligned in software. Because our cleaning tools aren’t finished yet, this was slightly cleaned in meshlab. Soft things like shoes don’t hold their shape when rotating, so the laces moved and needed to be deleted.

That also accounts for the missing data under the shoe laces in one place.  While not absolutely perfect, I'm very impressed that multiple scans on flexible shoe laces was able to deliver this result.  I'm guessing these are shoes in actual use.  Had they been prop shoes, this issue might have been able to be fixed using Super Glue to ensure the laces could not move.  This isn't a scanner's problem.  This is a target's problem.

THE SEAM STITCHING

The one area in which I do see some issues would be with some of the seam stitching.

Image - Show Stitching

THREE Scan - Shoe Stitching

 

The stitching where the toe and sides are joined was captured perfectly.  But, the stitching right above that, along the side of the laces and around the opening are less well defined.  Even so, they CAN be seen when we click on the above image to bring it to full size.  I believe that the stitches, themselves, are black.  And, black is notoriously hard to capture.  I don't know if this is a clue as to how well the THREE handles Black or not.  But, it is something to note as we make our observations.  But, In this case, it seems to matter very little.  The stitching WAS able to be captured.  It just is not as pronounced as the original.

THE UNDER SOLE AND SCAN SEAMS.

I am going to go back to my original capture of my examination in Meshmolder for this one.

 I COULD NOT FIND ANY SCAN SEAMS!

Remember, what we are looking at is a mesh created by combining 14 different scans.  And, the result appears to be seamless.  The MIGHT be a hint of a seam where the underside of the sole and sides of the sole meet. But, I don't think so.

And, the sole, itself, is wonderfully captured!  The detail is very, very encouraging for those of us potential;y looking to upgrade our scanning capabilities from a company we admire and trust.

I hope looking over my shoulder as I do my own "Due Diligence" in trying to assess the Matter and Form THREE is as useful to you as the process has been for me.

Fun with Scans

By the way, one of the reasons I use Meshmolder to evaluate the mesh is that it is a sculpting application.  First, it allows me to examine and manipulate the mesh in a variety of ways.   And, secondly, I want to be sure that output from the THREE is suitable for mash-ups and sculpting manipulation.  Here is my vertex view in Meshmolder.

Download this image and zoom in to see the quality of the vertex mesh.

The developer f MeshMolder also has a very low cost product called Meshrender that is a LOT of fun.  I couldn't help myself.  I had to see how the THREE 3D Scanner scan looked in Meshrender as a light blue plastic shoe.  Here is the result:

Meshrender - Light Blue Plastic (THREE Scan)

The surprising thing is that the image is MIRRORED in Meshrender.  I don't know if it's something I did, reorienting the mesh, or a natural Meshrender behavior.  In any case, it does not matter.  I simply flipped the image to correct the orientation.

Meshrender Output Flipped

As we move forward, I hope to experiment with Meshmolder and Meshrender with future test scans.  Scanning offers us a LOT of potential.

ADDENDUM:

I found out what I did wrong when I brought the shoe scan into Meshrender the first time and re-oriented it.  Here is a new render that did NOT mirror the mesh.  Cinderella meets Star Wars with a glass tennis shoe in space.

Matter & Form Scanned Shoe - Meshrender

Obviously, I am easily amused!