This post is as much about initiating discussion as it is about my own thoughts on the matter, so feel free to comment away, here, or twitter, or google+ or wherever – I’m interested in thoughts and perspectives.
So… 3D PDFs. They seem like a really good idea right? You can encapsulate 3D information to share with people. Stephan Lautenschlager wrote a workflow on producing them, and in it he stated:
“The ability to integrate three-dimensional (3D) models, which can be interactively manipulated, into portable document format (PDF) documents not only considerably improves their accessibility, but also represents an innovative, but so far neglected, approach for the presentation and communication of digital data… …As both authors and readers benefit greatly from their usage, it is argued that 3D PDFs should become an accepted standard in palaeontological publications of three-dimensional models.”
Now, I want to stress that I’m not having a go at Stephan, he’s absolutely right. What I want to discuss is in what he’s right with: “Presentation and communication of digital data.” The problem with that, is that it’s not “sharing digital data.”
3D PDFs fill a weird kind of niche – they contain all the 3D information, and allow users to manipulate that 3D data to view it however they wish. Obviously this is vastly superior to a video of a rotating model, it lets you tilt the object, zoom, and really look at everything you want to see, including things the author of the model didn’t think you’d want to see.
But that’s just teasing!
That’s showing the data without letting the reader actually get in there and use it. As far as I can see, there are two main advantages to supplying data as 3D PDFs:
- People can easily view it without specialist 3D software. (The ‘improves accessibility’ point by Lautenshlager)
- The data cannot be easily used by other people before you’re willing to share the raw 3D data.
Unfortunately, I’d say there are some pretty strong arguments against both these points. Point 1 is good in theory, but in reality a) people who are interested in the 3D model probably have, or are competent enough to get, meshlab, or something similar. B) Not all PDF readers are equal, and a lot of the non-adobe readers can’t open 3D PDFs (for instance, Adobe Reader Touch and Windows Reader can’t, I know I’ve had issues with Foxit and some Android PDF readers too, I’d be interested if all or some mac/ipad readers can), and this just exacerbates the problem, because now we have people for whom the 3D PDF just doesn’t work – if they’re lucky they’ll get a static image. The one time I would say that this point is valid is in complex scenes that can’t be exported as a single 3D file: visualizations of flow, for instance, can’t be reduced to a single 3D file for people to open in meshlab, but benefit from being viewed from any angle. Or perhaps virtual dissections, in which parts can be turned on or off with buttons (arguably, each part could be provided as a separate 3D file and opened together as layers in a 3D program like meshlab).
As to point 2, well, the arguments have been well trodden and discussed about sharing data and how important it is. ‘Hiding’ it in plain sight in a 3D PDF is a bit cheeky really, I think.
Except… There are ways to get 3D data out of a 3D PDF. They aren’t that common or easy yet, but it can be done. It can be as simple as having a specialist piece of CAD software that can natively open PDF files as raw 3D data.
And so we get to the part where I’d like your input:
If someone publishes a 3D PDF, is it fair game to take that published data, extract the 3D model, and use it in your own studies (properly cited of course!)? In discussions with colleagues, we likened it to extracting data from a published graph. But is the 3D model raw data, or is it a final figure? Does the journal own the copyright to the formatted 3D PDF submitted as supplemental data, or the data within it (surely it’s just the former, much as an unformatted manuscript remains the copyright of the author?)?
One has to ask of such data, why did the authors provide it as a PDF in the first place? It seems far easier to provide such data as OBJ or PLY files (or STL/PTS/XYZ/WRL etc etc) uploaded as supplemental data, perhaps in dryad or figshare. What did the authors feel was the benefit of making a 3D PDF? If it was genuinely to aid accessibility, then presumably they would have no beef whatsoever about another researcher extracting the 3D model. But what if they chose to produce a 3D PDF purely to avoid handing out the raw data?
Maybe I’m underestimating the accessibility angle. Maybe a lot of people do have Adobe reader, but don’t have any 3D software on their computer. My opinion is that there really isn’t any benefit to presenting palaeontological data like digitized bones and skeletons in a 3D PDF, as opposed to just uploading the raw files* and possibly a spinny animation.
The slippery slope
Having thought about and maybe responded to the above, I’d now like to posit to you a continuation of the theme. Advances in computer vision are enabling construction of 3D models from video, if someone uploads a spinning animation of a digitized specimen it could conceivably be reverse engineered [Indeed, this is a little non-palaeo side project I’m tinkering with]… Once again, I ask, is this fair game? Is the data considered published and re-usable? Do we need specific permissions from the original author? If so, why are we treating data in new media (3D models, animations) as somehow different to data in a graph or table?
I don’t know. It’s an exciting time to be working in palaeontology, what with all these new methods to generate, analyze, and communicate data. It’s something of an adaptive radiation of methods, if you will. But a side-effect is that we’re left wondering what is acceptable, expected, or necessary? As with all such explosions of accessibility in the digital world, there are two options; we can either try to crack down and control dissemination of data, or we can go with the flow; not just accepting it but actively encouraging it.
*(In the interests of doing as I say, I did so for my photogrammetry paper, and my Kinect paper, and have post-publication uploaded raw data for some tracks).
I agree that 3D PDFs are one of those things that are much better in the concept than in the practice. Working across several platforms (mostly Linux and the Android flavor of mobile OS, but also occasionally iOS), a 3D PDF usually doesn’t work for me. I suppose I’m just a “fringe” user, but I do think that the fact 3D PDFs do _not_ work well on most mobile devices is a major one.
Speaking to the issues of why more OBJ/PLY files aren’t available, I suspect that in some cases it stems from museum collections policies (or at least perceived policies that provide a convenient excuse). If I borrow a specimen and scan it, the museum may freak out if I put it online in a 3D printable form. That said, I’ve had reasonably good luck getting permission to do so (e.g., the Aquilops holotype is posted with surface scans–many of the crew from its museum were involved, so that probably helped–as is the baby Parasaurolophus we described a few years back–the specimen is at my museum, but is from BLM lands, so we got special permission to post the scan files).
Of course, I have a strong disagreement with the idea that museums are going to lose out on some massive revenue stream if printable files are out there (only the most exceptional specimens have any sort of cast marketing potential, and in any case if we want to go on about fossils are world heritage it seems a little dicey to start monetizing them in any form). I’ve discussed this in a few places on my blogs, notably here.
And as for extracting the 3D data from the PDF and using it…that’s the sort of thing that makes me a little uneasy (even if I think the idea of locking up the 3D data in an effectively unusable format is ridiculous). It probably depends on the use…but at the same time my gut instinct (as you note in the article) is that 3D PDFs are probably sometimes used as a way to obscure distribution of the “real” data (i.e., 3D morphology) for any number of reasons (e.g., museum policies, author recalcitrance, whatever). Depending on the case, it might fall in the line of “technically feasible but ethically dubious,” especially if it results in a use that is beyond what the original permissions for the specimen were (e.g., if I reverse engineer a 3D PDF to produce a print for an exhibit). I can’t quite pin my finger on it, but it just seems a little uncomfortable for me.
I think this is the kind of thought process that led to me writing the post in the first place. I want to believe that “if it’s out there it’s fair game”, but if it’s not supplied as raw data then why is it out there in the first place? I tried to tackle this at the end of the post, but I don’t think I was specific or indeed committed enough – I think we need to throw caution to the wind and just embrace it. See my reply to Daniel below.
I think that, at least in the case of rare or conspicuous specimens, the reason for not sharing the 3D files is rapid prototyping. Valuable specimens can be useful trading tools for institutions and research groups (say, for exchanging casts of rare specimens with other institutions) or a reason to visit the institution/museum because of the exclusiveness of the specimen.
When giving away the raw 3D data at full resolution, the fossil can be quickly printed by anybody willing to pay for it. Cheap, low resolution copies may begin to circulate in exhibitions, potentially diminishing the appeal of the original for many visitors: if that fossil has copies of it around everywhere, only few will care to visit the original. That could mean fewer visitors, which is something museums don’t want. Same with the trading tool value: if everybody can print his own copy of the fossil, why make an exchange at all?
In my opinion, 3D scans of fossils should be managed by the institutions that are in charge of the original fossils, not the journals.
As for reverse engineering… it happens as with traditional casting: creating a 3D model from a previous one: the result will not be as good as the original 3D model (as the errors in the digitizing process, though small, are still cumulative).
I think you raise legitimate, good, concerns. But it’s a little like the music industry when MP3s became A Thing. You either say ‘people shouldn’t take this data’ (and then watch powerless as they do anyway), or you try to foster a culture where the crappy casts are merely advertising for seeing the real thing. No one cares if I have an MP3 album of the rolling stones, but if I see them Live, it’s a real substantive thing (or at least it is to my dad!).
I’m one of those people that quite likes seeing casts in museums, but I know the value in seeing an original is far greater, (as you say, a replication always loses something). Maybe I’m projecting, but I think the public do too, and that’s something that can be tapped into.
Nice post, Peter. I totally think it is “fair game” to turn 3D pdfs into OBJs or any other format. Open data are open data; a 3D pdf does not come with a clause that you cannot use it in any other way. Of course it’s nice to ask the researchers, and might lead to new insights into the data/metadata and limitations etc., or new collaborations, but we can’t always invite everyone to do everything together.
Personally I don’t love 3D pdfs either but I use them sometimes to share image data, as some people prefer them over OBJs etc. There is still a luddite factor- I know some scientists who, with some coaxing, can get comfy with using a 3D pdf because a PDF is familiar and does not scare them, whereas they’d be intimidated to try Meshlab etc. with OBJs or other formats. However, with file space for data sharing becoming less and less of a problem (e.g. Figshare; Dryad; various specialist 3D data repositories like Morphomuseum), one can always share the same info in multiple formats; e.g. 3D pdf and OBJ etc. For the short term at least, maybe the shotgun approach like that is best.
I think for a figure that warrants 3-dimensions, a 3D PDF is a quick way to display the information to the widest audience. I also think that for those readers that want to study the figure in greater depth (including the Peter Falkingham’s of the world willing to reverse engineer), providing OBJs/STLs/ect would be good practice. I agree with John. At the moment, providing both is probably the most effective approach.
Another advantage of 3D pdf is that you can share an ‘object’ with multiple parts (*.obj files). This gives you flexibility in the way to interact with the data—maybe you want the surface structures ‘on’, perhaps you’d like to turn them ‘off’ and view the deeper structures, or render the surface structures transparent. I agree with Andy, that Adobe’s implementation of 3D pdf is not as ubiquitous as we’d like (no help for iPads—but it works fine in Acrobat for linux). Moreover, Adobe has somewhat orphaned the 3D pdf format. Myself, I’m struggling to find a presenter app that allows me to share 3D data that are comprised of multiple ‘objects’ (*.obj, *.ply, or whatever). John makes some good points that, like it or not, at the moment 3D pdf works for the larger audience. Yes, I am looking forward to the day when we can share an agglomeration of 3D objects (preferably as a single file) and know that most of our audience has the means (apps) to view them. I appeal to all to let us know if a good alternative to Adobe Acrobat Reader and the 3D pdf is emergent.
Meshlab project files allow combining several geometries, too.
Our world is still split into two parts: the one in which people are able to directly load a PLY/STL/OBJ and the one in which people are stymied by the extension and cannot get it to show. There’s Meshlab to fix that, but until everyone has it 3D PDFs are helpful for bridging the gap.
That said, it IS nice to have a way of annotating 3D models 😉