A 1.525Mb PDF of this article as it appeared in the magazine complete with images is available by clicking HERE
The Sense hand-held 3D scanner from 3D Systems was recently evaluated by LIDAR News. We set-up some quick, basic tests, tried our hands at scanning various objects/ people and evaluated the output models in a few different 3D software packages. Finally, we report on the efficacy of this low-cost, consumer-grade scanning unit.
The Sense
The Sense scanner is a small, portable, light-weight handheld device that was brought to market by 3D Systems late last year. Developed by Israeli manufacturer PrimeSense (maker of the Kinect and recently acquired by Apple), it went on sale (for $399 USD) via internet orders in the fall of 2013 and has subsequently become available at Staples, albeit at a slightly higher price. The device is distributed through the Cubify "hub" a retail site that, in the words listed at www.cubify.com,
"…offers everything from cocreation with favorite brands, a curated shopping experience centered around personalized fashion, dcor, toys and more, a community inspiring the latest in 3D design and printing possibilities, and access to the easiest home 3D printers and related products."
It is fairly easy to deduce that if 3D Systems (who acquired Geomagic– makers of 3D software and scanners) are going to be able to sell lots of 3D consumer-grade printers and thermoplastic consumables (such as PLA and ABS), then there needs to be a device at the front-end feeding the pipeline, providing "content-to-print" The Sense . was conceived as a means for the average user to scan objects (and people) of interest for replication by the consumergrade 3D printers sold through Cubify.
3D Systems have been making different types of sensors, some of which are similar to the Sense but are priced much higher. The technology is still essentially the same: a depth sensor and 2 stereo cameras. 3D Systems refer to the technology as "physical photography" The Sense does not have . the accuracy of the commercial-grade professional scanners sold by 3D Systems, nor a device like a coordinate measuring machine (CMM), but at 1mm, the Sense is (presumably) accurate enough to provide reasonable fidelity when reproducing smaller objects. The depth resolution is acceptable for larger objects such as a guitar, bicycle or humans.
Testing
In evaluating the capabilities of the Sense, we approached it as novices and run-of-the-mill consumers. We planned on using several test objects that we felt represented an average range of possible scanning scenarios a typical user would face. What we didn’t plan on was the sheer number of attempts we would need before finally obtaining mediocre 3D models.
The first attempts with small objects (see Figure 2.) involved walking around a table with the objects in the center. Virtually all scans failed, primarily due to loss of tracking. Even when the tracking was re-acquired, the resultant models were more like amorphous blobs (see the teapot and hat in Figure 3.). The second difficulty encountered was maintaining the required range. The Sense software offers a set of concentric "target rings" that turn blue when the scanner is placed within an acceptable distance to the objects/subjects, and turn red when the range is either too close or too far. To ensure a tight fixed-range was nearly impossible with small objects.
We then tried fixing the Sense to a tripod and move the objects on a microwave turntable. This ended-up producing pretty good results (toy truck in Figure 3.). Although, it is possible that the Sense software may need to "see" the rest of the room move relative to the fixed object being scanned.
Similar to the small object scanning, we struggled with walking around human subjects, but were able to finally obtain "water tight" models by having the subject sit on a swivel stool (Figure 4.) and slowly turn 360, collecting data on just the head, neck and shoulders. Full body scans were problematic. We were unable to adequately maintain tracking. It seemed as if the tracking was lost each time as soon as we moved the scanner to 180 around the subject–to the back half of their body (This also occurred with smaller objects.).
Software
Along with the Sense data collection GUI, we also used Cubify’s Sculpt (see Figure 4.), which is a model design and manipulation package (based on original products from Geomagic) available at additional cost. Sculpt has a clean interface, but required substantial processing power. It has many interesting features for the "artistic" users out there, people who are interested in creating mash-ups of objects or altering models from scans in an artistic way. We had an easier time with an open source, free package called CloudCompare, which didn’t appear to require the horsepower that Sculpt does and it was quicker to use it to manipulate and edit the models (see Figure 3).
The Sense software settings do allow for limited "tweaking". These were adjusted multiple times for the tests, but we didn’t realize a substantial improvement in being able to produce decent scans or time-to-model. There are likely configurations that we didn’t find that enhance performance. This is an area to possibly pursue in further testing.
What users have to say
While perusing the Internet for instructional videos and commentary by other users of the Sense, it became apparent that there is a consistent frustration with using the scanner as well as with the results produced by it. Both professional testers (e.g. LinkedIn Laser Scanning Group, 3Ders.org, Ben Heck) and novices alike described significant struggles to achieve good results. As with other users, our experience was different from the depiction in the marketing videos of users doing a quick 30 second scan of a human body, rendering a seamless mesh as a final product. Time and again, with both large and small objects, we were unable to obtain results that were close to those promised in the advertising. In almost all cases, the device lost track (See Figure 5.) of the object/subject. It took incredible effort just to maintain "tracking", and many, many attempts before we succeeded.
Competing (emergent) technologies
In concept, the Sense is a brilliant idea. It is intended to bring together simplicity of use with reliable 3D scanning. The difficulties in putting the scanner into practice notwithstanding, the Sense has substantially advanced the ability of novice and even experienced users to collect 3D data. However, there is an emergent threat in the form of "reality capture" via photography and also from smart phone/tablet-attached scanners. For instance, 3D Systems just recently announced the iSense adapter for iPads. [In our tests, it would have been much easier to maintain tracking of the object if there had been a way to mount the Sense to the notebook to see the range circles and maintain tracking] They may cannibalize sales of the Sense, but this way they can saturate (across platforms) the consumer market with content generators for 3D printers.
Reality capture software will also threaten the active scanning systems like the Sense. Using ordinary passive imagery capture devices like cameras found on tablets, smart phones and point-and-shoot devices, users will now be able to render 3D models using relatively inexpensive or free software such as Autodesk 123D Catch.
Conclusion
There is a technological and economical gap that exists between scanning and 3D printing. The Sense is clearly targeted at this gap. For 3D printing to take off among the "masses" there needs to be , inexpensive and consumer-obtainable data to be available for creating 3D prints ("content-to-print"). The question is whether the Sense will fill this role. From our tests it was plain to see that the difficulty to scan anything quickly and easily (without glitches) is a serious impediment to universal uptake by average consumers.
The Sense can be a fun "toy" to experiment with 3D reality capture. It is great for learning about 3D scanning and the science behind three dimensional data capture and analysis. The Sense is also a fun gadget to use as a data capture device for direct input to a 3D printer. However, the accuracy realized through the scanner is not enough for copying small parts and objects (for something like reverse engineering). Additionally, the effort required to produce a "decent" result is much too great for it to be used in any sort of production environment.
Perhaps in the hands of more skilled professionals, the Sense can produce great results, but for the average consumer it presents a challenge and doesn’t live up to the billing. Buying at a retail store like Staples at least affords the novice user the opportunity to return the Sense since Cubify does not accept returns.
Special thanks to Anne Gutelius for being a willing and patient "subject".
Bill Gutelius is the President and co-founder of Active Imaging Systems (AIS). In 2007 he formed AIS where he consults for commercial and government clients on active and passive imaging technologies and their applications.
Sidebar:
Fast Facts
$399 US list price
Based on PrimeSense (Kinect) Technology
Class I IR laser
Real-time output
PLY, STL, OBJ output files
Quick Specs
Spatial x/y resolution 0.9mm @0.5m
Depth resolution 1mm @0.5m
FOV: Horizontal: 45, Vertical: 57.5, Diagonal: 69
Operating range: Min: 0.35m Max: 3m
Scan volume Min: 0.2m x 0.2m x 0.2m Max: 3m x 3m x 3m
Sidebar:
Pros
Very inexpensive
Generic output files
Lightweight/Portable
Fast data collection
Includes color imagery Fully integrated with Cubify modeling software and 3D printers
Clean user interface
Cons
Very tricky to obtain good results
Output is not always high fidelity
Does not work very well with small objects (<10inches)
Software has limited functionality
Needs to be tethered to a Windows device
Loses tracking easily
Narrow range (to target) settings
Tethered via USB to computer/tablet
A 1.525Mb PDF of this article as it appeared in the magazine complete with images is available by clicking HERE