WORD ON OWNERSHIP, COPYRIGHT AND FAIR USE
Very little has been written or said in the paleontological sphere about the
ownership and copyright of 3D data (but see Mallison and Wings, 2014: 8).
However, this is of critical importance for users of online 3D data, lest
you create legal problems for yourself or your institution. Anyone who has
worked with museum specimens knows that they often have a web of ownership
and permissions that extends from the collection manager to the board of
trustees. This can lead to a situation where someone who may have collected
and prepared the specimen does not ultimately own the rights to the images
of the specimen that they may have taken. 3D would logically follow this
same web of ownership, with the result that any 3D models of specimens would
still be owned by the museum or repository. However, the museums or
repositories are often bound by the requirements of granting agencies or
university requirements that stipulate open access to data generated. The
United States National Science Foundation has been helping to lead this
charge with the expectation that data generated with its funds be made
publically available to other researchers when reasonable (https://www.nsf.gov/bfa/dias/poli
The first step in this process is to check what the website housing the data
universal code regarding 3D data in academia. Each website has its own take
on it (some do not even mention copyright at all). AfricanFossils.org,
Morphosource, and Phenome10k all put their data under Creative Commons
Licensing (https://creativecommons.org/licenses/by-nc/4.0/). This means that
as long as you properly attribute the source of the data, you can use it for
any non-commercial purpose, like education or research. Other websites have
their own, unique statements. The Digital Morphology Museum (KUPRI) simply
states on its homepage: “The goal of this site is to enable you to view the
scans of non-human primates and mammals and to download scan data from our
database for your original research. It is our great pleasure if you can
make use of these data and we hope that they will provide new insights into
primate and mammalian evolution.” Aves3D endorses usage for ‘personal
education of website visitors and non-commercial scientific research and
education’ as long as the source is acknowledged. Digimorph’s statement
endorses use of their data for ‘personal education’, but is silent on the
subject of classroom usage. We recommend that when there is doubt about how
you can use the data, contact the site administrator for guidance /
permission. You may need to pursue further permission from the person who
scanned the item.
What you need to begin
To start with, you need a decent graphics computer with lots of free memory.
SPIERSedit recommends a 64-bit system with at least 2Gb of RAM, a fast/large
hard-disk. SPIERSview recommends a fast graphics card. On that computer, we
recommend the following software packages (freeware): ImageJ (http://imagej.nih.gov/),
SPIERSedit and SPIERSview (http://spiers-software.org/; Sutton et al.,
2012), MeshLab (http://meshlab.sourceforge.net/), netfabb Basic (http://www.netfabb.com/basic.php),
and software supplied by your 3D printer (i.e. Makerbot Desktop for the
Makerbot printers, http://www.makerbot.com/desktop). This final portion of
the protocol was written specifically for an extrusion-type 3D printer—it
can easily be modified for other types of printers.
you find scans?
The online repositories for 3D data are just developing now. Several have
emerged in the past few years, and there is no centralized search mechanism
for them yet. There are still scientists who post CT-scan datasets on their
lab websites (e.g. Dr. Larry Witmer, http://www.ohio.edu/people/witmerl/lab.htm),
or places that have them available for viewing, like Digimorph (http://digimorph.org/).
Datadryad.org, a site for the storage of scientific data, hosts some
CT-scans, and we expect that number to grow in the coming years. There are a
host of sites specific to 3D models themselves—where you do not have to
worry about converting CT-scan movies into models—places like Morphosource,
Phenome10K and others (Table 1).
Table 1 -
Web resources for 3D skeletal models. 3D indicates that 3D filetypes (.stl,
.ply, etc) are available. Dollar signs ($) mean that the models are not
movies or 3D files is fairly straightforward on most sites. On DigiMorph, it
is a little more complicated, as many of the CT- scan datasets there do not
have download links. The way to download them is to click on the slice movie
name, in order to bring up the pop- up movie window. What you do next
depends on what internet browser you are using. It is easiest to download
the movie files (.mp4) using Mozilla Firefox as your browser. Using Firefox,
you right-click on the movie and choose Save video as.
After you have downloaded your movie, you need to do some initial processing
in ImageJ. To load the movie in ImageJ, you need to go to File > Open, then
choose your movie. A popup window will appear, in which you will need to
select Convert to 8-bit grayscale and Use virtual stack. The first thing you
should do is to remove any text in the movie. Some movie files have slice
numbers or text (file name, scale bar, etc.) in them, and you need to remove
them from the files to make a printable file. You can remove them later
using the SPIERS software, but it is easier to remove them in ImageJ. Just
use the rectangular selection tool and erase the text from all frames.
The most important thing that you need to do is to convert the file from a
movie to a folder that contains all of the ‘slices’ of the model.
Essentially, what you are doing is creating a digital ‘flipbook’, with all
of the digital ‘pages’ of the ‘flipbook’ in one folder in the proper order.
To do this, you need to go to File > Save As > Image Sequence. This will
cause a popup window to appear; in the window you need to choose the format
.bmp and change the name / numbering as appropriate. After you finish with
that window, it will bring up the Save window, where you will choose where
to save the files. We strongly recommend creating a new folder in which to
save all of the new images.
To create the model you need to load all of the images into SPIERSedit, and
tell the program how to recognize which pixels in each image are to be a
part of the resulting model. SPIERS has been discussed in publications (e.g.
Sutton et al., 2012, 2014) and interested readers are urged to consult them
and the manuals for both SPIERSedit and SPIERSview.
First, you open SPIERSedit. You load your images by going to File > New—and
then select all bmp files in your folder. Next, you will use the Generation
window to set a grayscale value that captures the bone or fossil to the
degree you want (basically, you are now telling the program which values of
gray will become a part of your model). This approach works best for
isolated specimens; it is a bit trickier when the specimen is surrounded by
matrix (a fossil) or tissue (a preserved modern animal cadaver). Usually, if
there is matrix or tissue surrounding the item of interest, it will usually
be a darker shade of grey. Thus, by refining the greyscale level, you can
digitally ‘prepare’ a fossil from the matrix or ‘dissect’ a bone from an
animal. First, select a slice of the animal that encompasses a lot of
variation. Then, use the sliding bar to select the grayscale value—click
‘generate’ to see if the value works well. If not, change the grayscale
value, and hit generate again. Each model is different, and a guess and test
approach must be used. Once you are satisfied with the grayscale, select all
the slices in the slice selector window, and click generate again. This will
apply your grayscale threshold to all slices. At this point, you can go
ahead and make a model that can be 3D printed.
To make a model, you go to output > export SPIERS view and launch. At this
point you may notice some problems with the model—for example, the model
being compressed or extended unusually in the axis that each slice was
produced. This is because the thickness of each slice of the model is not
set correctly. To fix that, go to the Output window, click on the settings
tab, and change the value of slices/mm—usually to 0.5, 1.5, or 2 slices/mm
as these are typical values for the number of slices/mm. The correctness of
the values used can be crosschecked against the dimensions of the specimen.
You may also notice that the model is ‘holey’ or unusually thick—these are
signs that you need to re-adjust the contrast in the model.
SPIERS exports its models as a .spv file which can only be read by the
program SPIERSview. Within SPIERSview, you can apply some smoothing to the
model, as well as export it as a .stl file, which is the file type that will
be used in all succeeding pieces of software.
Dissecting the model
Certain types of models will require additional preparation. For example, if
you have a scan of a cranium and jaw together, you may want to digitally
dissect the jaw from the cranium so that they can be articulated with a
movable joint as in life. Having a fully articulated jaw and skull with
mobile jaw joint is obviously more realistic. However, dissecting the jaw
from the model you have produced is difficult. It can be done during the
post-processing phase (i.e. after the model is printed), but this runs the
risk of breaking your model beyond easy repair (especially for those who are
not gifted sculptors). To dissect the jaw off the model you need to go into
curves mode in SPIERSedit—the use of which is covered in detail in the
SPIERSedit manual. The curves option is a way to essentially lasso an area
that you want to include in a separate model. Simply put, you lasso the
lower jaw in each slice and combine them together to make a model of the
lower jaw. You can use the same approach to remove extraneous text / scale
bars—basically, you select everything but extraneous text on the frames, and
then copy that curve to each frame.
For the lower jaw, create a separate curve for the jaw, and use that curve
to create a boundary between the skull and jaw. Difficulties will include
interlocking teeth and the jaw joints. Add extra points to the curve to make
a more complex polygon that will enable you to weave the curve around the
interlocking teeth, or the jaw joint. Once you have created a curve that
includes all of the points you want, you need to move all of those points to
a mask. You must be in mask mode, click the toggle button to ensure that
your curve is filled in rather than just an outline, create a mask, select
both your new mask and your curve (and select all slices), and the go to
Mask > Mask from curve. Then click on your new mask, and open the output
window, and create a new output object from your mask.
Post-processing of the digital model
Despite having the finished model, you are not quite ready to print. Open
the model in Meshlab, and you will notice that the model appears very dark.
Due to an idiosyncracy of SPIERSedit, the models are essentially created
inside out. To fix this, you need to go to Filters > Normals, Curvatures,
and Orientation > Invert Faces Orientation. Other useful tools under the
Filters tab include tools for simplification (i.e. making the model smaller,
e.g. Quadriatic edge collapse decimation), and for smoothing (e.g. Laplacian
smooth, which for each vertex a new position is calculated based on the
neighboring vertices). See http://www.3d-coform.eu/index.php/tools/ meshlab
for more information, as well as tutorials on youtube.com. See Sutton et
al., (2014) for more details. These techniques are essentially removing data
from your model; when used to too great of a degree, they can easily reduce
the resolution and make the model appear blocky (there are no hard and fast
rules for how much smoothing or simplification you can do to a model). Use
them with caution in an iterative fashion and make sure you save your model
before you use them.
Netfabb Basic is a good program to do some basic repairs on the model. Go to
Extras > Repair part to fix minor errors in the model. You can also use
Netfabb Basic to slice up models that are too big for the print area of your
printer (be sure to check the dimensions of the printable area of your 3D
printer). The pieces can then be joined together after printing to create a
larger model than could be printing a scaled down version of the model by
itself. To do this, you must use the Cuts panel in the lower right hand
corner of the Netfabb Basic window. We recommend trying to cut the object
along a plane of symmetry. One advantage of this is that you can have the
cut surface placed downward on the 3D printer and avoid printing many of the
extraneous supports that would normally print under a more irregular surface
(extrusion based printers have to build thin supports in order to support
overhanging bits of morphology while the plastic is still molten). After
digitally cutting the model, you must export each part individually as an .stl
There are several different ways of printing out 3D models (Tschopp and
Dzemski, 2012). One of the most commercially accessible methods is
extrusion-based additive printers (like the Makerbot printer which was used
by the authors). In this mode of printing, molten plastic filaments are
extruded onto a surface, layer-by-layer. The model is built of these
successive layers—which can be printed solid, hollow, or with a
honeycomb-like internal framework for strength. Externally, the printer
will print a network of supports to support overhanging parts of the model
(Figure 1A). Once the printing is finished, these external supports will
need to be removed. The supports will, in most cases, break away easily,
leaving little trace of their presence on the model. Others will need to be
delicately removed by sanding, or with a hand-held Dremel tool. We recommend
lightly sanding the entire model with a Dremel tool (with an abrasive wheel
attachment) to help minimize the layered appearance.
Figure 1 -
Tasmanian Wolf (Thylacinus cynocephalus) cranium from the Idaho
Virtualization Lab (https://sketchfab.com/models/67f1975ef2e940929b4eee
f3de769f6e#). The skull took about 12 hours to print total. A) Freshly
printed skull halves showing the extraneous supports that were printed along
with the skull. B) Sanding half of the skull to aid in fitting it together
with the other half.
Once you have the parts cleaned of support material, the assembly and
finishing process can begin (Figure 2A). First, you should dry fit all parts
for best fit and alignment. Get rid of any expectation of a perfect
fit—commercial grade 3D printers will typically introduce some amount of
minor warping or distortion. All flat surfaces that are to be mated should
be sanded smooth to ensure closest contact (Figure 3A). A sheet of 150 and
320 grit emery (sanding) paper spray mounted to a glass plate makes an even
sanding surface. This will accommodate most printed pieces. The flat side of
the piece is rubbed in light circular or figure eight pattern against the
abrasive surface, 150 grit first then 320 grit (Figure 1B).
When the surfaces are satisfactorily smooth and fit cleanly they may be
glued together (Figures 2A-C, 3B). Start by drilling several holes in each
surface being glued (Figure 3C). 1/16 to 1/8 in holes are preferred
depending on the surface area. The holes need not match up from side to side
as their function is to afford an additional anchor spot for the glue. Check
that no burs from drilling the holes have altered the fit of the pieces. The
glue of choice is 5 minuet epoxy, it is strong, permanent, and sets quickly.
We prefer Devcon brand but any good quality brand should suffice. Mix the
epoxy according to directions and apply evenly to one surface. Quickly mate
to the other surface and rub them together to squeeze out as much excess
epoxy as possible. T h i s w i l l a l s o f o r c e the glue into the holes
where it will form a “rivet like” connection (Figure 3C). Wipe away all
excess epoxy from the joint, a cloth dampened with denatured alcohol may
make this easier. Hold the pieces in position until the epoxy is set
(approx. 10 minutes). Repeat this process until all the pieces are assembled
to complete the skull and /or jaw (Figures 2B-C).
Figure 2 -
Progression of assembly for a mosasaur skull and jaws (Tylosaurus proriger
[https://sketchfab.com/models/5b8ef5d8a2ce4ee986b645bb672fcc20], no longer
available for download). The model was constructed out of 12 parts, and each
part took between 12-24 hours to print. A) The cleaned and sanded pieces,
ready for gluing. B) First step of gluing together completed. C) Final
gluing completed. D) Skull after covering with epoxy and painting.
Now that the skull is assembled you will want to fill all joint lines and
imperfections (Figure 3D). Several materials will work for this process.
Selection of which is best to use is left to your budget and preference.
We recommend the following: PC-11 Marine white epoxy paste. This is nice for
small imperfections and joint lines. It has a good working time, is
reasonably inexpensive, and smooths easily with a finger dipped in alcohol.
Aves Studio’s Apoxie sculpt and Apoxie paste. The sculpt material is a two
part plastic clay that’s great for major fills and additions. The paste is
more fluid and can fill and smooth large rough areas. Both have a long work
time and smooth easily with Aves safety solvent or alcohol. Devcon 5 minute
epoxy Gel is nice for small quick imperfections and joining pieces that have
a troublesome or gapped fit.
All of the above mentioned materials can be cleaned up with denatured
alcohol or lacquer thinner and can be sanded when cured.
After addressing the imperfections and joint lines you will want to smooth
the skull surface to at least partially hide the isomorphic lines left by
the printing process. Smooth-On Inc. makes a product called XTC-3D that is
specifically designed to do this job. It is a two-part epoxy with good
self-leveling properties and a short work time. We have found it best to mix
small batches of around 20 grams by weight or ¾ of a plastic shot glass
full. Follow the directions and pour the mixture into a shallow tray from
which to work. This will slow the cure process and maximize your work time.
Clean your brush with lacquer thinner or acetone before the product sets and
cures. Two or more coats may be needed or desired to eliminate the
isomorphic lines. XTC-3D can be colored with SO Strong, another Smooth-On
product. This can be useful to see where you have been and provide a base
color. XTC-3D cures to a high gloss finish. Though we have not tried it, we
believe any good quality epoxy finishing resin might work as well.
The final stage will be painting the skull model. We recommend several light
coats of a flat white primer such as Krylon to provide a uniform base color.
This also makes it easier to spot any drips or other imperfections you may
have missed. When all corrections are made and you’re happy with the surface
apply whatever color treatment you prefer. We tend to use acrylic paints as
they generally come in a wider range of colors from which to work. No matter
what treatment you choose; if the skulls are going to be handled, you may
want to protect them with several clear topcoats (Figure 2D).
We estimate that our department has spent about $500 for these models (now
numbering over 40)—plastic filament for the printer, paint, epoxy,
etc.—versus the over $10,000 it would have cost to purchase casts and
replicas. This is almost certainly a poor estimate because many of the
specimens we have printed are not commercially available, or not
commercially available at the sizes we have printed them. Printing large
models can take up to a week for a large, complex piece—multipiece print
jobs will take even longer. The finishing of the models can take several
days, including time for sanding, gluing (if necessary for multipiece
models), applying epoxy, and painting. As an example, the Tasmanian wolf
skull (Figure 1) cost less $10 in materials and could have been finished
(print-to-paint) in a week. In contrast, the Tylosaurus model (Figure 2) was
printed over the course of a month and cost about
$100 in filament alone.
Figure 3 -
A detailed schematic drawing showing how to glue together pieces of a
multipart model shown as a progression of four steps, A-D. A) Step 1,
showing the poor fit of the two parts prior to sanding. B) Step 2, sanding
the surface of contact improves the fit of the two pieces. C) Step 3, by
drilling holes into surface of contact, one can improve the grip of the
epoxy glue. D) Step 4, by filling the external joint with epoxy type paste,
one can make the joint essentially invisible.
discussion here has thus far been focused on vertebrate fossils and skeletal
elements. One can just as easily print invertebrate fossils using the same
methods above. A word of caution:
any specimen with delicate bones (many types of fish) or thin, spindly legs
(arthropods) will be very difficult to make. They are difficult not because
of printing but because removing the supports that print as a consequence of
extrusion printing is difficult to do without also inadvertently breaking
the fragile model.
We acknowledge valuable support and advice from Roger Runquist (WIU) and
Russell Garwood (University of Manchester). We thank the developers of the
SPIERS software suite, Netfabb, and Meshlab for making their products freely
available. This contribution was greatly improved by constructive reviews
from Stephen Lautenschlager and an anonymous reviewer.