Hands On With ReconstructMe

If you've ever been shopping for a 3D scanner, watch out for the price tag. Many commercial units providing high quality results and convenient features are well beyond the pocketbook of almost every hobbyist, upwards of USD$50K in some cases. 

 

Are there less expensive methods of 3D scanning? Several 3D Scanning kits or hobby versions exist, but there are compromises in function and you'll still be short hundreds or even thousands from your wallet. Meanwhile, there is one way to begin 3D scanning and it's possible it may cost you absolutely nothing! 

 

That method is ReconstructMe, a software tool we've written about previously. Today we've been hands on experimenting with the tool to see how it works.

 

ReconstructMe is software that uses the commonly-available Microsoft Kinect 3D sensor to produce full 360-degree 3D scans. While other Kinect software exists, the ones we've seen tend to take a still 3D image from one location. ReconstructMe is quite different in that  it lets you image the backside of the object as well as the front. In fact, you can scan all sides continuously to produce a true, full 3D model. 

 

We took time in our lab to test ReconstructMe to experience its installation and use. 

 

We found the worst part of the ReconstructMe experience to be software installation. The tool requires several drivers to be installed as well as a Microsoft framework. If that doesn't tell you something right away, we'll say it: ReconstructMe runs only on Windows machines.

 

If you run Linux or OSX you don't have a chance. Even using virtual Windows is not acceptable, as virtual machines frequently don't forward the required information as expected by ReconstructMe. We found only one example where someone had successfully installed ReconstructMe on OSX using specific virtual machine software, but even then it did not work well due to performance limitations. 

 

Our advice: read the installation documentation very carefully. Then read the documentation again very carefully and follow it precisely.

 

ReconstructMe really needs a good video card to work well because the software attempts to use the processing power of your graphics card to handle the scan data to produce the 3D model in real time. It is possible to use ReconstructMe without the graphics processing boost by scanning first and producing the model later, but this is awkward as you're never sure your scan truly covered the entire object as you can't see what you're scanning. ReconstructMe provides a handy list of graphic cards to check compatibility and performance. 

 

One problem we encountered with the installation had to do with software drivers. Each time we plugged in the Microsoft Kinect sensor, Windows would conveniently install the wrong drivers for us automatically. We had to rapidly cancel this frustrating auto-installation to enable a manual installation of the correct drivers. 

 

Once you've successfully installed the software, using ReconstructMe is relatively easy, even though it uses a command line interface that would likely frighten non-technical users. A single command launches ReconstructMe and you're presented with a two-pane window with one view being the live camera and the other the 3D model under construction.  

 

Performing the 3D scan could not be simpler. You simply sweep the camera view around your object (which must be contained within a one meter cube facing the sensor when it is activated) and gradually complete the scan. You must look for "holes" in the generated model and repeat area scans to fill them in. Smooth, slow motions seem to collect the best 3D information. If you (as we often did) move the sensor too close to the subject for the Kinect to resolve (400mm), you'll receive an annoying BEEP. Or three. Too many beeps seemed to freeze the scan, but you could still save your results at that point. 

 

An alternative approach is to place your object (or person in our case) on a rotating platform (or office swivel chair as we did) and slowly rotate them in front of the stationary Kinect sensor. ReconstructMe scoops up the data and magically produces the 3D model in front of your eyes. While this approach seems easier, you still have to pick up the sensor to manually scan areas not seen by the sensor while rotating (the top, for example). 

 

The quality of the scans is relatively good. It's hard to measure this, but our experimenters thought the accuracy was in the 3-5mm range. Perhaps this could be increased by more careful and slower scanning. In any case, it's good enough to grab a 3D model of a person without keeping them longer than they'd become annoyed. It is not good enough for anything small, however. Jewelry scans would be impossible with ReconstructMe, as we suspect the Kinect is simply unable to resolve such details. The Kinect doesn't scan clear objects as it seems to see right through them. Similarly, it gets confused by complex fine details such as fuzzy hair. 

  

ReconstructMe can save the generated 3D model in several formats including OBJ and STL. Our experiment involved STL, which was produced correctly. We found the scans to include stray data, suggesting you should be careful where you place your subject - avoid nearby items. Be careful not to capture stray arms from an assistant rotating the swivel chair, as we did in this image. There were a few small holes in the model, but these were very easily corrected with 3D model management software such as NetFabb. More than likely you'll scan more object than you need; the table or chair your object rests upon will also show up in the scan. This necessitates some 3D editing to remove unwanted portions and close off the bottom. 

 

ReconstructMe is available in two forms: Commercial and Non-Commercial. The non-commercial license is available at no charge, meaning if you already have a USB-capable Microsoft Kinect, this could cost you absolutely nothing to try! On the other hand, if you require a commercial license, you'll need to pay €360, or about USD$450. 

 

For rudimentary 3D scanning of large items this price can't be beat, even considering the installation complexities and user interface peculiarities. We recommend you give ReconstructMe a test today. 

  

Via ReconstructMe

Hands On With the PP3DP Up! 3D Printer

Meet Phil Letourneau, an Up! 3D Printer Operator. This week Phil dropped by the Fabbaloo office to demonstrate the rarely seen Up! 3D Printer. We put this elusive device put through some basic tests to see how it works. The result? Overall, we're pretty impressed - but there are some caveats. Here's what we found:

 

The device retails today for USD$2690, but Phil managed to grab one of the few early shipments that were discounted to USD$1500. He's not sure if he'd buy it again at the full price, given the other 3D printer options available today. You can choose a pre-assembled unit from any of six colors; Phil's was bright orange.

 

The metal-framed Up! comes mostly assembled, and includes a pretty extensive toolkit, which Phil describes as "Well Equipped". The kit includes a multi-tool wood cutting kit, excellent for removing the voluminous support material that seems to appear on almost every print; snips for cutting filament; a really thin metal lifter; and a set of cheap industrial gloves.

 

Phil says he's paying PP3DP approximately USD$35 for a spool of 0.7Kg filament (which, by the way, is a lot thinner than standard RepRap/MakerBot 3mm filament).

 

The Up! includes an unusual filament feeder mechanism: the filament travels through a tube, which tightens up as filament is eaten by the print head. When the tube tightens close enough to the spooler, it touches a mechanical switch to engage the spooler motor, which then delivers a few centimetres of fresh filament. It's fascinating to watch this process repeat, for some unexplainable reason.

 

The Windows-only software provided by PP3DP is both good and bad. Good because it tremendously simplifies the process for the user, who is faced with a straightforward control window with only a few knobs, as opposed to RepRap/MakerBot/RapMan SkeinForge users who must diddle with hundreds of parameters. Up! operators merely choose things like "Solid" or "Loose", etc. While this simplification is good, much control is lost. For example, there doesn't seem to be a way to NOT print a raft. Also, if a print doesn't come out right, there's often not a lot you can do about it. However, Phil said PP3DP has been very good about releasing upgraded software on a pretty frequent basis, often with interesting new features.

 

A great feature is automatic support structures. You simply open up any STL file and the Up! does its best to print it. Very often we observed support structures printed, sometimes even when there was no apparent need for them. We spent a lot of time peeling prints. Interestingly, one problematic STL file, unprintable on a MakerBot, actually printed successfully on the Up!

 

The quality of the prints is superb, particularly when the printer is set to its finest resolution: 0.2 (we suspect that means millimetres, but it's not completely clear, as with most aspects of this overly mysterious machine). The software offers four choices of resolution, with the "coarsest" being perhaps not as good as the average MakerBot print. The poor resolution prints also take mercifully less time than the finest, which seem to go on forever.

 

Curiously, the warm-up process involves a massive test extrusion - of about half a meter of filament, or so it seemed. The only reason for this, we thought, might be to encourage additional filament sales. We've also heard of challenges in using filament from places other than PP3DP, since you have to somehow get the foreign filament onto the Up! spool, but it can be done, apparently. 

 

The purchasing process was interesting: Operator Phil had to make arrangements to wire funds directly to a previously-unheard of factory business somewhere deep in China. We asked if it was uncomfortable to send that amount of cash to an unknown company far away, and Phil agreed, saying "it seemed totally sketch".

 

The Up! seems to have a couple of power challenges. It needs power for the heater, electronics and a few motors for XYZ axes and extruder. Where does it get this power? From the THREE separate power plugs you must insert into several nearby electrical outlets. Worse, the device has NO Off switch. That's correct - to turn it completely off you must unplug THREE cords.

  

Noise factors are both good and bad. The Up! is astonishingly less noisy than a MakerBot; you can hardly notice it printing, as it's almost silent. However, the silence is broken by extremely loud and annoying BEEPs that must signify something of importance.

 

Bottom Line: This is a pretty decent device, producing high-quality prints with little effort required by the operator. We believe this device could be successfully operated by someone with only light technical skills. However, the simplification of the interface doesn't quite hide all the complexity behind the scenes. 

 

Via PP3DP

Test Parts

Television has the famous colored Test Pattern. 2D printers have "Test Pages". But what should 3D printers use as the gold standard for testing and comparing output objects?

 

With the recent marketing comparisons of 3D Systems and Objet technology, and now a post from Open3DP, it's seems logical that the industry should develop defacto standard test objects. Open3DP needed to develop a consistent object so that their experiments with various powders could be effective. Who else would need such objects?

 

We believe manufacturers should use standard test objects. Consider the scenario where you're shopping for a printer and you are sent identical test parts printed on each of the candidate printers. You'd be able to perform visual and tactile direct comparison tests. 

 

The standard test object should likely include features that enable comparison of surface quality, sizing accuracy, resolution on all axes, edges, fragility, color and other dimensions. 

 

Via Open3DP

A Resolution for Resolution

Last week we enjoyed reading Tim Pickup's excellent review of several contemporary 3D printers. But hidden inside the article was something that we found intriguing.

Tim pointed out a chart that compared the 3D print resolutions of various 3D printers. As you can see in the chart, there are vast differences in print quality between the devices.

Of course, resolution is not the only matter of concern when buying a 3D printer. However, this is one of the first direct comparisons we've seen. Like a test pattern, we can see this type of comparison becoming a standardized evaluation for 3D printers in the future. But there are questions:

What should be the form of a resolution test?
How does color alter your perception of the resolution?
How fine and detailed should the test model be?
How can the test's results be properly represented in 2D form (for publication)?

Time will tell. We forsee a standardized resolution test for 3D printers that will be commonly used by users and manufacturers.

Via HalfPricePrototypes