1. Building 3D content
to last
Stephen Gray
Associate
JISC Digital Media
info@jiscdigitalmedia.ac.uk
2. Content
• JISC Digital Media 3D work so far
• Culture of data reuse/challenge of reusing 3D
• 3D file formats
• Significant properties of 3D
• Metadata for 3D content
• Work yet to be done/conclusions
3. www.jiscdigitalmedia.ac.uk
• Support for JISC Advance HE/FE/GLAM
• Video, sounds, images & 3D
• Website: „good practice guides‟
• Mailing list, blog and Twitter
• Helpdesk service
4. JISC Digital Media
3D Toolkit
• An Introduction to 3D Digitisation
• IPR for 3D Digitisation (with JISC Legal)
• Sharing 3D Content
• Publishing 3D Content Online
• Seminars, workshops & case studies
• Examples of 3D in research...
9. Data reuse requirements
(government, research and heritage)
• Data „availability‟ mandate from RCUK (2013)
• Publishers (e.g. BMJ) data requirements
• data.gov and more „open data‟ EU directives likely
• HLF „digital outputs‟ available for 10 years, usable
for 5 years, post-award
10. Some reuse expectations
Digital content should typically be:
• In a suitable format to support reuse
• Adequately described to support
reuse
• Available/reusable for x years
11. Culture of reuse:
A 3D problem?
‘Standard’ digital content Digital 3D content
Long-lived file formats clear
Significant properties defined
Metadata standards in place
Long-lived file formats unclear
Significant properties undefined
Metadata standards not in place
13. File formats
“We are often inhibited... by the large numbers
of proprietary formats used to analyse, process
and visualise 3D data.”
- Jenny Mitcham, ADS 2012
14. • Widespread adoption (proprietary often
beats open source)
• Minimal „dependencies‟ (open often wins)
• Code in public realm (open or „semi-open‟)
• As simple as possible
‘Standard’ file format
selection criteria
15. 3D file formats: the state of play
• Over 166 different 3D file formats in common usage
• All are complex (define Geometry, Appearance, Scene
properties, Animation)
• Excluding proprietary/undisclosed formats many open
formats exist (e.g. Meshlab outputs: PLY, STL, OFF,
OBJ, 3DS, VRML 2.0, U3D, X3D and Collada)
• Some even have „sub-formats‟ (e.g. DAZ Collada &
Autodesk Collada)
• All are different (e.g. Geometry stored as parametric
surface, faceted mesh, pointcloud, boundary only or solid
geometry)
www.turbosquid.com, a solution in practice:
16.
17. Common 3D file formats
• .dae (Collada) XML-based format is
gaining ground
• .3ds and .obj most commonly used
as exchange formats
• .obj is currently still 3D „workhorse‟
18. .obj still popular*
Data Description Preservation format (Archiving)
Required?
Original raw scan file ASCII TXT and Native Yes
Pre-mesh point cloud ASCII TXT Strongly advised if
archiving a derived
Polygonal mesh
Polygonal Mesh OBJ + MTL (ASCII) Required if final
dataset
Decimated Mesh OBJ + MTL (ASCII) Required if final
dataset
*From ADS/AHDS laser scanning good practice guide
http://guides.archaeologydataservice.ac.uk/g2gp/LaserScan_3-1
19. • Image and Spatial Data Analysis Group (ISDA)
Conversion Software Registry
– Repository of info on conversion software
– Complements „file format registries‟ (e.g. PRONOM)
• ISDA‟s own software tools for 3D conversion at
http://isda.ncsa.illinois.edu/drupal/
3D formats: useful tools
20. Q. Format.x seems best, will it last?
A. No
All file formats will become obsolete and will need to
be updated over time.
We must ensure 3D content continues to „perform‟ in
the same ways, independent of format.
21. Significant Properties
• Updating/converting formats = information loss
• Important to ensure files still „perform‟ in the same
way as before
• Files should convey the essence of a digital record
• Files akin to a musical score (rather than a
gramophone record)
22. Significant Properties
• Files „do‟ lots of things, what do we
really care about?
• Defined as significant properties
• Ensure these are maintained,
regardless of current or future file
type
• Documented as metadata...
24. Metadata
• Actual data is the 3D model, metadata describes the
3D model
• Searchable, can be queried used for browsing (and
managing) digital content
• 3D-specific standards slowly emerging
• Back to www.turbosquid.com...
26. Metadata
Metadata for 3D Models (Boeykens, Bogani 2011)
suggest the following are also described:
• Are any materials used physically correct?
• Are advanced shaders/plug-ins used?
• Are layers or components present?
• Is lighting/physically correct lighting used?
• List any external files required
27. Metadata
The POCUS* project suggests we also describe:
• Audience and type of interaction expected (i.e. who is the
model aimed at and how can they interact with it?)
• Methods and techniques used to create a model
(hardware model numbers, software versions etc.)
• Datasets used to create a 3D model (e.g. laser scanning,
geophysics, lidar)
• Flag when features of a model based on estimation
rather than evidence
*The preservation of complex objects (POCOS) Vol.1,
Visualisations and simulations, 2012
29. Work yet to be done
• Significant properties of 3D should be communally
defined (extend INSPECT work?) to save
duplication of effort
• A shared (if basic) 3D technical metadata standard
could be defined and adopted by repositories
• The shortage of 3D digitisation training with a public
sector focus should be addressed
30. Conclusion
The sharing of 3D content is complex. It would
ideally be considered (and costed) before any
content is created.
Archiving is far easier when repositories (museums,
archives, universities or similar) work alongside the
creators of 3D content with preservation as the
common goal.
- paraphrasing ADS
This slide outlines the structure of the presentation.
The initial guides we’ll be releasing in the area of 3D content will be:An Introduction to 3D DigitisationIPR for 3D Digitisation (with JISC Legal)Sharing 3D ContentPublishing 3D Content OnlineWe also hope to offer seminars, workshops and ongoing case studies as our support in this area grows.But now let’s define what we’re talking about when we say ‘3D’ pr ‘3D content’ with some real world examples.
Developers are working with veterinary scientists to produce Augmented Reality displays which visualize how the skeleton of an animal relates to its external body structure, as well as to compare the skeletons of different animals and explore their differences and similarities. (http://augmentedrealitydevelopmentlab.com/modules/#Animal%20Skeletons%20Revealed)
Motion capture (MOCAP) is being used within the arts and sports science to create accurate representations of how people move. Commercial systems are already on the market which allow us to develop that perfect golf swing! (e.g. www.qualisys.com/applications/biomechanics/sport-science/)
Rapid prototyping or ‘3D printing’ is now fairly common but pioneering users are often found within engineering disciplines. Astoundingly some ‘printed’ machines don’t even require assemble – they’re created ‘pre-constructed’! (e.g. www.thelookingglassclub.com/#/print-your-own-walking-machine/4551977074)
Putting aside the special uses of 3D for a moment, let’s look at some at one significant aspect it shares with other data types. Some major agencies are now encouraging or even insisting that, where public money was used to develop it, data is made freely available.
We could spend the entire morning listing who these organisations and institutions are, but suffice it to say the UK government, most research funders and a substantial number of heritage sector funders fall into this category.
Common requests are being made of the creators of publically-funded data. That data should be:Created in a suitable format to support reuseAdequately described to support secondary usersAvailable and/orreusable for x number of years
This is usually achievable, if often challenging, for the creators of what we might call ‘traditional’ digital media. We know which file formats are likely to last, what it is about a file or dataset that is important to preserve and how these things are usually described.Now let’s return to 3D content – we don’t yet actual ‘know’ any of these things; which file formats will survive, what a 3D model will be used for in the future nor how bets to describe it so that other users can understand what we’ve created.
Let’s look in more depth at these ‘uniquely 3D’ challenges.
Firstly – which file format do we select if we wish to share our 3D content? The situation is summarised well by Jenny Micham of the Archaeological Data Service – there are a vast number of choices and many are owned and mentioned by sole commercial vendors.
Common criteria do exist which can guide us when selecting (for example) a format for a digital photograph. Is the format widely used? Does it need only a few, non-specialised ‘supporting technologies’ to be in place before we can view it? Is it an ‘open’ format and is it fairly simple to use? If the answer is ‘yes’ to all of these, then we might decide the format is a good one to use.
But now let’s look form the perspective of 3D content – the first criteria doesn’t help us much because there over 166 3D formats in ‘common use’. All of them are complex (i.e. none are ‘simple to use’) and many require equally complex technologies to be in place before they can be accessed.Even the ‘chose open’ mantra doesn’t help us much as there are many open formats and some even have sub-formats, which makes the situation even more complex. There is no single contender set to become the ‘universal’ 3D open format of choice. One reason for this is that many of the formats work in radically different ways: geometry might be stored as parametric surface, faceted mesh, pointcloud, boundary only or solid geometry.And yet 3D content is created, used, shared and archived everyday so some pragmatic, de facto standards must be in use. And they are – let’s look at the commercial Turbosquid site where thousands of 3D models are bought and sold every day.
We can see that, or all the formats which might have been selected to represent this elegant coffee pot only two were selected. One of these may well be very familiar to everyone here - .obj. It’s remained popular for some time because it works.
We’re happy to say that our research shows that the open formats are the most popular and while the Collada is gaining some ground, .3ds and the old favourite .obj are still the most widely used 3D file formats.
This is further backed up by severalpublic sector archives, who request .obj for practical reasons. It’s easy to read and preserve. Here we see the formats recommended by the Archaeology Data Service. You’ll notice that .obj is not recommended in isolation, but is association with ASCII TXT ‘snap shots’ taken as various points in a model’s development.
The Image and Spatial Data Analysis Group maintain a list of software capable of converting between 3D file formats, as well as providing their own software conversion tool. If you’re stuck with a less-than-ideal 3D format, the ISDA site can most likely help you.
Now Id’ like to speak briefly about the concept of ‘significant properties’. We know that no single file format will last forever. Technology will advance and all formats will become obsolete eventually. Therefore, we must plan to convert anything of value to a new format at some point in the future.
However, when we do that some things will change, we might say that some properties will be lost. What’s important is that our 3D content continues to do whatever it did before (help us perform that perfect golf swing for example). We need to see our files as a musical score rather than a gramophone record.
We need to list the things a 3D model ‘does’ and ensure it continues to do those things regardless of current of future file format. To use an example from text documents: we may find out via a survey that our users simple read the documents we give them, in which case we may wish to convert all our documents to .pdf format to make publication more convenient. On the other hand we might find out that our users both read and make changes to the documents we give them. In that case we would not convert to .pdf as the ability to make changes would be lost.Once we know what they are we document significant properties within an item’s metadata record.
I’d now like to speak briefly about metadata and 3D content. What is metadata?
Metadata is ‘cataloguing information’. Our 3D model is our data, the description of that model is metadata.Why is it important? Because without a description we couldn’t find the model. Also, managing a collection of models would be very difficult. In a similar way to 3D file formats, best practice is slowly emerging, meanwhile we have de facto standards. Let’s return to Turbosquid.
Here we see a very technically complete (and very useful) set of metadata ‘element’ for 3D content, drawn form the Turbosquid site. Notice too, that we’re also given a clear indication of what we’re allowed to do with the model (licensing information) and an informal ‘rating’ provided by other users.
A very good paper by Boeykens, Bogani ‘Metadata for 3D models’ seems to largely agree with Turbosquid’s approach. In addition though, the paper also suggest that the following pieces of information be provided along with a 3D model, the final bullet suggest we list any dependencies the 3D model might have which are not provided along with it.
Similarly, The Preservation of Complex Objects project suggest a few extra metadata fields which could help ‘round-out’ a technical record like that used by Turbosquid. These are particularly relevant for 3D models produced for research purposes.
I’d like to conclude with a look at what’s yet to be done in this area.
Anyone can make use of the ‘significant properties’ approach but we gain extra benefit if we do so as a community. While the Jisc-funded INSPECT project defined the common significant properties of certain data types it did not do so for 3D content. Any project which did so would be well received.As individuals or individual institutions, we might also create own metadata standards in order to describe 3D content. But if we do so in isolation we risk not being able to easily share with other institutions. Therefore a shred approach to developing a basic and easy to use 3D metadata standard would also be very welcome.If we consider sharing and preservation when 3D content is being created, much of what follows becomes easier. This is especially true for 3D digitisation techniques such as laser scanning where the dangers of a commercial solutions (rather than one built on open tools) can be particularly acute. Therefore a 3D digitisation training provider is needed, preferably one based within the public sector.
My final slide once again draws from the ADS who have done some great work in the field of 3D archiving:Sharing should be considered before any content is created, is should also be properly funded, not just ‘tacked on’ to other activities. The second ADS point is for archives, repositories or libraries. Such organisations, working alongside 3D content creators, play a pivotal role by helping ensure that the three-dimensional documents of tomorrow are preserved as well as the two-dimensional ones of yesterday.