Jump to content
Ultimaker Community of 3D Printing Experts
ultiarjan

Fuel3D scanify

Recommended Posts

After a long wait I got my Fuel3d scanner.

I've tried to create some sample situations that should work, with structured backgrounds of carped or rice, but I'm really struggling with my Fuel3d scanify.

 

 

>>> Edit; I got a replacement unit (from , thanks) and this one is ok. The "Failure in 3D reconstruction" errors as reported here we're due to a hardware defect <<<

 

 

If you have one, please share your experience here.

 

My last +/- 20 scans 100% end up in a "Failure in 3D reconstruction" message.

If you ignore this you can get a scan, but quality is poor.

 

Reported by Fuel3D it has issues with

 

-Low Texture Surface

-Reflective Surfaces

-Cavities

-Geometric Features

-Overlapping surfaces of different layer heights

-Protrusions

-Large objects

-Small objects

-Transparent Material

 

So it ain't easy finding a suitable test object :)

Have a look at these publications to see some examples of objects that will give issues;

https://www.dropbox.com/sh/a8yoj6xb698tanu/AAAcbuGS2HA6MDsQ8f3wvTIba?dl=0

 

I basically have not been able to get any good scan.

All scans result in "Failure in 3D reconstruction", so I really wonder if my hardware is defect. The support guy mentionend "your camera may have fallen out of calibration" but It's not clear to me if this is a hardware or software issue (waiting for a reply).

 

After the failure you can proceed with the scan after some warnings;

F1 FailureInReconstruction

F2 oops

F3 Continue

The issue with "protrusions" is very visible on a face scan, the sofware will add a surface under a nose, whether you scan straight or from below. Also a nose's dimensions are of, on a frontal scan the nose will stick out less than IRL making it hard to stitch a frontal scan of a face with a side scan. At least for now as the current (beta) stitching software doesn't seem to modify the individual scans, but merely positions them.

C1 noseLow

C2 nose scanStraightFront

Besides the scanning difficulty there's other stuff like the software won't run on my win8 laptop, so I'm stuck to my win7 desktop (though I did see it work on a win8 tablet).

 

Here's a few of my test examples;

Art object with texture, and a textured background.

The background (textured carpet) is really warping and distorting.

A1 beeld

A2 beeldFail

A3 WavyBackground

Bed of rice, simplest test also results in a "Failure in 3D reconstruction"

B BedOfRice

Piece of tree trunk, about 10cm diameter, 15 cm high.

Seems even here the depth difference is too much to cover.

boom1 fail

Safe to say I'm fairly unhappy with the unit at the moment, but I hope the expected software update (promised for Q2) will make a difference.

Last point, the marketing of Fuel3D is still very unclear.

They now offer 3 different levels of software with the unit.

euro 1299,- starter pack

euro 1599,- plus pack

euro 3149,- studio advanced

 

It unclear whats inside the sets, but my guess is that "starter" is only scan + stitch and "plus" will include mesh modifying software (there's a coop with Uformia MeshUp).

 

I've been told that the early buyers (I ordered mine around october last year) get the "plus pack" software as compensation for the long delivery time.

 

Share this post


Link to post
Share on other sites

With my limited experience with 3D scanning, I know about some problems,

transparent and reflective models won't scan properly, the reflection always mess up the camera or sensor,

there are special white powder spray cans especially for this purpose.

scanning smaller object scans could be because of the limited resolution and lens deformation, and the scanner is unable to stitch the scanned sides together, most scanners have a ideal distance to size ratio, that makes for ideal scans.

I know that with scanning smooth surfaces you need a rough surface next to it, for the scanner to have a common point to stitch the scans together, and very structured objects can have to many unknown points to get a good scan stitched.

also dark and light colors can really mess up depth perception of the scanner, and a even color surface is better, again that's why the powder spray cans are there for.

you still need to find the ideal way to scan objects and have to learn what you can and cannot scan, try some pillows, they are more uniform in color and have not to complex shapes.

stay away from transparent and reflective surfaces.

http://3d-scanning-tips.blogspot.nl/

 

Share this post


Link to post
Share on other sites

damn....

My fuel3D scanner is shipping early April. I was hoping that it was going to allow me to scan to my hearts intent, but from what you posted, I'm going to have my work cut out for me :(

Also, I had no idea that I would have to pay for software? That sucks.

Keep us posted

 

Share this post


Link to post
Share on other sites

These are most valuable posts. If you see my topic on the ISense scanner, I have to agree that the resolution achieved is not really quite adequate. A friend has ordered the fuel3d scanner so I will be able to try it out.

Share this post


Link to post
Share on other sites

I visited Fuel3d offices last October to look at the scanner and get sample scans for processing. I came away with several .stl files one of which had 700,000 faces in the mesh, so plenty of resolution if your software can handle it.

At the time they were uncertain which package was going to contain which level of software, although I did get the impression that the mesh stitching software would not be in the base level. This is important if you want to take 360 degree scans, maybe even 180 degree scans. The unit is not a scanner, it is a camera. To get a 360 scan you will need to take at least four pictures and stitch them together.

 

Share this post


Link to post
Share on other sites

I was just testing a design for a tablet mount, in a very quiet room, and when turning the scanner it made a sound, I'm pretty sure there is something loos inside around the arena of the bottom camera. This may explain my issues.

I'll check on Monday how to get it replaced.

 

Share this post


Link to post
Share on other sites

I got our Fuel3D a week back and have made a lot of scans, we love the unit and got some very good face scans out of the device. I was also frustrated with the software, but now use MeshMixer (free beta) to refine the print after the processing. It's not a quick process, but it does work.

 

Share this post


Link to post
Share on other sites

I got our Fuel3D a week back and have made a lot of scans, we love the unit and got some very good face scans out of the device. I was also frustrated with the software, but now use MeshMixer (free beta) to refine the print after the processing. It's not a quick process, but it does work.

 

very interested, can you send me some links, etc...?

thanks,

-Pat

 

Share this post


Link to post
Share on other sites

Don't use the sprays for this scanner. It wil not work! The fuel scanner uses features from the object. If it cant find (and match) features, it is unable to make a 3D reconstruction.

The fuel basicly combines two types of scanning; stereoscopic (eg; the same as what 123 catch does, it matches recognisable points and triangulates them) and photometric stereo (Light object from multiple angles, the way light changes tells you something about the orientation of the surface, but not about the depth).

The innovation is in the combination of these two data sets; The stereoscopic method is rather error prone and very sparse and is unable to really find surface normals. The other is able to do that, but unable to find a depth. You can quite cleary see artifacts that are explained by this; The nose bridge is an area that is integrated based on normal field, using the depth image as a scaffold.

The calibration of such devices is in principle a mathematical model that maps the camera focal length and their distance / rotation with respect to the other. If you change the focus / lens / orientation of the cameras, you need to calibrate again. They probably hardcoded the calibration, so you are screwed i guess. Calibrating the camera's yourself isn't that hard, provided you have acces to the software.

It will probably help if you give the software a 'lot to work with' by scanning objects placed on a mat with recognisable points (QR/AR codes work quite well for this)

 

Share this post


Link to post
Share on other sites

The support guy from Fuel3D indeed told me that if the unit is "out of calibration" it has to be send back to them, Apparently this should normally not be necessary as the hardware is fixed.

My unit clearly has a hardware issue (there's something loose inside), luckily I bought it via a retailer (makerpoint) and I can pick up a replacement this Thursday ;)

 

Share this post


Link to post
Share on other sites

I put your pic into Photoshop to try and enlarge it to see the facets but it is too small. It looks pretty smooth, what resolution is that scan, i.e. no of facets/faces. Also have you tried stitching the other part of the face to that model?; interested to know how that part of the software is working.

 

Share this post


Link to post
Share on other sites

Oh, also have you tried cropping yet with the Fuel software? When I visited them they cropped the files for me but that seemed to be the cause of the meshes having 800-1200 errors in the mesh. I did accept that the software was still under development so I am wondering if that has now been cured?

 

Share this post


Link to post
Share on other sites

Just picked up a replacement scanner for my bad unit. Did a few test scans... all scans are ok, so I can confirm my bad results where indeed a hardware issue. Will post some results later.

Thanks @Makerpoint for the replacement service!

 

Share this post


Link to post
Share on other sites

I think it's a nice piece of hardware, will show some results in the coming days/week, but a bit busy atm.

I did a quick test and Meshlab is indeed pretty easy, and atm seems to provide better results than the Fuel3d beta sticher (did only 1 test), but ofcourse this is just a beta so I'm expecting it to get a lot better.

The unit certainly has it's limitations, but I don't think there's anything better on the market for this price...

I also got a dell venue pro 8 tablet this week, and the scanner works fine with it (2 Gb memory, windows 8.1) so now I have a portable unit. Will share the mounting bracket I made later.

@Solid, do you have any specific need for the scanner? just let me know if you want me to scan/test anything specific...

 

Share this post


Link to post
Share on other sites

Some impressions of my test experience.

 

ART WORK.

The scanner has issues with many objects, as described above and in the Fuel3D documents. To give a few more examples.

 

testSetup

rotary

I made a test setup, scanner on tripod, rotary table for controlled angle steps (used a rotary disk of a nodal ninja panoramic photography set I have). Unfortunately using fixed angles is not really useful, as you will have to check each scan for failures anyhow, and sometimes scan from a little different angle.

 

Hole Protrusion

for example holes and protrusions are very difficult. I tested with a little artwork (high school art-class of my daughter).

As you see in the photo above the top hole changed into a protruding shape, and the bottom hole is strangely fading out. Adjusting the scanner to a slightly different angle improves the result.

 

protrudingBottom

another example of an error produced is the bottom part of the object. The foot is moving inward in reality, but outward on the scan.

 

Trick

for a complete scan you'll also need a top scan. If you just scan the top it will produce a useless result. If you use a trick, like adding a "carton collar" it's possible to create the scan.

 

 

Hear2

Hair1

differenceindepth

 

HUMAN HEAD.

Scanning a head is in itself very simple, best done with a windows tablet mounted to the scanner. The issue is you need many many many many scans, and the waiting time between scans is 30 seconds. Nobody will be able to hold still for a long time. Next to this the accuracy of the result is low, for example the protrusion of a nose will be less when scanned frontal, some scans of the back of the head will show strange "corners" (see pictures). This makes it all very difficult to stitch the individual scans. If any of you got good results please share them, I would love to see them....

 

 

TagPosition1

You need to position the tag carefully, it should not be obstructed, but also not to close to the edge of the image. In above example the distance to the edge is to close and the scan failed.

 

 

conclusion; the resolution is nice, the accuracy is poor.

Do NOT expect a simple quick point and shoot device (which the fuel3d marketing is trying to sell you).

In a controlled environment, with a fixed object, careful planning and checking each scan, you can get a good result for the few objects that meet the "scanability criteria".

 

Scanning a human head is very difficult. As it takes 30 seconds between each scan keeping the same position is impossible for the person scanned. To get any decent result you'll need to check each scan, and redo some, which is very difficult when using a mobile tablet setup, as the tablet (mine= win8, 2Gb) is very slow in processing the scans. In reallity you'll do a scan session, and redo some scans later. I have not yet successfully reproduced any human head (to the disappointment of my kids).

 

I think many issues are intrinsic to this scanning technique, but I do hope some will be overcome with software improvements. I hope the scanning will get faster. I'm really curious about the quality of the stitching software that should be released Q2, the current beta is not very useful IMHO. Regarding the stitching software; I saw on the fuel3d website the packaging of the box has been changed, and now no longer contains the text "includes stitching software" think this will mean that you'll have to buy the more expensive "plus" version to get the stitching software.

 

 

 

 

 

* disclaimer :-P

opinions based on first test scans, not on hundreds of scanning hours. software used was final software, stitching software is still in beta.

 

Share this post


Link to post
Share on other sites

Well Im not giving any positive nor negative advice ;) just sharing the unit has limmited use with regard to the type off objects you can scan... and it shure aint point and shoot...

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×

Important Information

Terms of Use Privacy Policy