21 January 2025 in Deventer

Questions & Answers Tijdlab (Notes Julie)

About the interviewees

• Brendon: gaming/ game design 3d modelling photogrammetry; game design is part of their core business (game creation and technology in Enschede

• Rob: Director of Tijdlab; presenting CH to the wider public; VR experiences, online games; archaeologist; best method for presenting hte data - a 3d scan, a complete vr environment; they developed archeolink (a database system used by gov orgs to maintain archaeological datasets); also have their own online 3d viewer, which is scientifically approved unlike sketchfab; domain experts + technicians make up their teams; also 3d imaging in universities

• Aleks was their first contact (TwinIt)

Their workflow

• How do you work? 

  • start from customer wishes/ask requirements to be met? Yes

• Where do you get your models from?

  • Normally they create all their own 3d content (scanning interiors, sculptures); having control creates the models and ensures the quality control

  • In their experience, if they don’t have control over the scanning, the 3d scans aren't of the quality that they can use in game design environment or printable - QUALITY CONTROL; you know what you’re working with

  • They also use The Unreal Store, Epic Games, Mega Scans.

    • NOTE: All of these have now moved to http://Fab.com ??

• What software do you use?

  • They combine multiple scans in the 3d scanner software itself; or later in Maya, Arctic, Blender, Maya Z-Brush (e.g. to fix a hole in the model)

  • There is not one software that does it all - it’s always a combo of multiple scanning and modelling software - for example Artec Studio, Reality Capture

• What file types?

  •  Obj, stl, xml for metadata

• “If we get a model from someone, we ask what they did it for, and check that it has the necessary elements for that use case

Questions about quality

• To what extent are you interested in metadata & paradata? 

  • Paradata is important to know how the object was made

    • type of scanner, when, by who “what kind of model do we have, how should we treat the model, and can we get source files!”

  • Metadata is important for creating experiences

    • E.g. for making sure it’s the right object for their game (if it’s 13th century churches, they don’t want one from the 15th century); to know what dates are associated with what parts of the church

  • In some cases, privacy restrictions could apply to some paradata

    • who scanned it, who worked as a designer on it - their names could be private info that providers/we might not want to show

• What makes a good model for you (and what doesn’t)?

  • “A perfect cube only has 6 surfaces, every surface is 2 triangles, so there should be 12 triangles; if there are 11 or 13 that’s a problem.”

  • Could others use your scans? Yes, anyone with 3D knowledge should directly be able to open it in an online viewer, modeling software, or printer software.

  • To define that standard would be quite interesting - they’d like updates from us. 

  • Scanning landscapes - the quality of the scans was good; in archeology the are used to working with small objects vs. landscapes the higher detail the lower the deviation will be 

• What is a good platform (for example, do they indicate licensing)? And how would you measure quality?

  • They don’t use libraries often, they scan their own objects. But in the event they do, licensing is important.

  • Viewers (sketchfab) 

• What are some common problems you encounter when trying to find or use 3D models?

  • An item might be “high quality” (over a million polygons) 32K texture - but making this enormous data set work on a 4k screen means you will lose data, which means you can lose actual important necessary data

  • Double data or double layers, the data stacks on each other which creates distortion - “it’s a fine line between too much data and too little data”

    • You get a feel for how much data is too much 

  • It’s more work to fix the scan (700 euro) than to scan (half that price)

Questions about TwinIt!

• What were the 4 most commonly encountered issues in the TwinIt! Models? 

  • The strategy when scanning - they focused on the wrong thing (the trench from a medieval castle) - the detail of the trench was very low because there was so much data in such a huge scan - so they missed the target, so the quality of that scan is ‘low’

  • Too much data is not necessarily helpful!

  • They need to look at the real questions - what should be the point? What are we actually scanning? What is the goal - what will it be used for?

• Were there models that you liked to work with, and if yes, why?

  • The Spanish jar

  • The church from Finland was very good

    • “The mesh itself was ok, but for gaming and rendering it was really good”

  • Some partners are more skilled in scanning  - capacity building opportunity?? 

    • They were finished game models were fit for purpose - they were generally good, they knew how to create a generally good model for an online web viewer and it worked for our purposes as well; they had clearly created scans for online viewers and modeling software

  • Latvian monument - the flowers show up only in the texture, not in the shape - because flowers are really difficult 

    • whoever brought it from point cloud to mesh knew what they were doing 

  • Slovenian library - too MANY textures (500 textures) - they were maybe inexperienced and thought that maximum texture settings in their scanner is the best, but that’s not the case (does the sound files section in EPF outline how to avoid excessive ‘noise’ in a recording? This might be similar)

• Worst models - portuguese building model that was really only a point cloud

• What steps did you take to improve the models?

• A previous example they worked with: Call of Duty scans of Amsterdam (over half a million budget - so everything was high detail) 

• Most they TwinIt! Items were downloaded from the website, some were sent directly, that didn’t have any influence on quality.

Questions about licensing

• Are you interested in licensing?

  • Only for models that aren’t important - from other platforms, they’ll get ‘filler models’ like fences, cars, trees to fill up the environment

  • Anything we scan for someone else, it’s theirs but please say ‘this scan was made by Tijdlab’

• What licenses do you prefer to use?

  • Open licenses

  • The libraries we use at least have filters to choose reusable licensing

• What licenses do you avoid?

  • Copyright

Paying for models/platforms

• Do you ever pay for models or platform subscriptions?

  • Rarely, maybe 5 euro for over 200 scans

• If yes, what percentage is paid vs. unpaid?

  • Very small percentage is paid

• Steam/paid libraries/free libraries (Steam already has many users)

  • They find models on The Unreal Store, Epic Games, Mega Scans

About Europeana

• What do they want from Europeana?

  • To see what they want to work with, and what they’re downloading (as users)

  • To work as a partner (we could jointly apply to grants/projects, we could use their viewer)

  • They don’t own any of the scans so they can’t be a provider, they just have the viewer and act as a tech partner; they are a full end-to-end technical partner 

• Would you go to Europeana in the future? - If no, when would you?

  • We reached out to them technically

• Preference for other platforms? (see other notes; also, their own viewer?)

• Downloading experience

  • Showing the size of the whole thing, also its indiv parts and allow user to select which they actually want to download

  • A small README file with this info as a minimum

  • (see Additional Notes below)

Tijdlab’s users

• Which users are reusing your content? How are they using it?

  • CH, gaming industry -http://witsenshipbuilding.com allows people to play the game online in the browser and to download the game and play 

  • We’d advise europeana to publish on STEAM bc it’s easier than publishing on our site and has a big user base (100 thousand ppl have found and downloaded the TwinIt STEAM game) - CHECK IT OUT; they also help with applying to grants (they know good funders)

• Which content is most popular with them?

  • Depends on their ‘research question’ - what they need, why they’re doing it (see notes above on this)

Additional notes

Idea: order tiers by use cases

1. Raw scan files or point clouds only (for research) - level of detail of the model itself doesn’t have to be high - raw files are more preferable than a nice render

2. For 3D printing (STL only for simple plastic printers, OBJ for printers that can handle texture)

3. For opening and exploring in 3D modeling environments; non-open license: 

   1. Rendering (exporting 2D images)

   2. Animation (specific topology for this different than other rendering use cases - look it up)

   3. Game environment, XR environment

   4. Maintenance of the monument itself (needs to be super high quality)

4. Same as Tier 3 but with open license

The ideal download

A better folder could contain the constituent parts (which could be very high quality and very usable individual) and also all of them stitched together

1. Are they all there? The pieces of the castle, the whole thing, the obj, the stl

2. What quality are they each at? Is it too big or too small? We have to set ranges for each scale of model (small objects, vs. machines, vs. landscapes/monuments)

Filtering in the Europeana UI

• Idea: filter by the level of expertise of the user itself  - some people can work with anything

• Also limited by the end-users hardware - printers, computers, mobile devices, web browsers

• The hardware used to make a scan, the way they process the data is like a black box - researchers don't know what algorithms are behind it, that’s why the most raw files are useful (because the data behind the model is better quality )

Tijdlab’s 3D viewer

• Storage is included in the viewer on the EU server(s), and the provider gets different amounts of storage based on their needs

• https://tijdlab-reseller.pedestal3d.com/gallery/  - works publicly

• If they worked with us, we’d get our own domain - would be something like  “europeana.pedestal”

Questions & Answers Tijdlab (Notes Lianne)

Participants:

• Brandon van Doorn

• Rob Van Haarlem 

• Julie Van Oyen

• Lianne Heslinga

Location: Deventer

Rob

Has founded company Tijdlab. Now 10 years old. Focussing on heritage, topic always in center, most important, then see how to present it. F. e. Holographic display or online games.

Created ArcheoLink, a database for archeology. 

Also created an online 3D viewer, that is for sale, annual fee. (link)

He got into contact with Europeana via Aleks, for TwinIt!

Brandon

Game designer, studied game creation & technology. Is specialized in 3D modelling.

Newest project: in Brugge, scanning to create 3D models, but also create user experience.

Using artec scanners.

Brandon: high resolution/quality vs. game engine.

We normally like to make the scans ourselves for the partners.

Too much data creates distortion.

There is a fine line between too few or too much data.

Following your own workflow, from scanning to UX is much easier (and normally cheaper) than fixing other people’s models.

Could others use your scans easily? - Yes, normally in obj/glb, very easy to load in any kind of 3D software. We did not define a standard/follow a standard, but it is a fingerspitzengefuhl. If we would define it a bit more, it would be very close to being a standard.

For example from the models that were delivered for TwinIt!, an archeological landscape model was good, but it was too much, more an entire 3D dataset, and it was an entire terrain, instead of the trench focus point, which would be the interesting part. Normally in archaeology, different levels of scans are made, for different purposes. 

For research, it is of course very helpful to have source files/raw scans and a lot of information, but just showing it in a virtual exhibition is another use case.

Before anything, we always ask our partner: ‘What is the aim of your project/scan?’ 

That will define for a large part how we will work. For example very detailed vs. aerial. For very detailed 3D digitization, keep the limitations in mind, also f.e. in terms of hardware. 

And determine the right output format. Stl for example has no color. 

For metadata, it is even possible to describe this for a part of the model. If you for example digitize a church, you can have metadata for just the tower, that is built in a different time than other parts. 

When necessary, we make use of photogrammetry and combine this later into the software. We use Artec 3D Studio.

If partners already have scans themselves, we ask: ‘Why did you scan it?’ You should ask this question to your providers as well. 

Then you can link quality to usability. Because different qualities are needed in different situations. 

You have to match quality against the usability question that the partners have asked themselves. 

Examples are:

• Printing

• Maintenance

• Publication

• Research

• Rendering

• Animation (is like rendering)

Sometimes cutting a model up can make the quality higher, if certain parts have higher quality than the composition. If the topology of the model itself is broken, the model is not good. Perhaps the raw scans were bad. 

It could be an idea to assign ranges for different objects/scales. If it is missing some file (f.e. texture) this could be an indication. 

Julie: we could work with filters. 

For expert users we could give them raw scans. This would then be a usability question for advanced users. Then we would make it dependent on hardware, level of expertise. 

Always indicate quality/size and what you need it for (= use case and range).

Software is a black box, you don’t know the algorithm, so no good for research. 

TwinIt! Models that were good/ok:

• The Spanish jar 

• The church from Finland 

These were both finished gaming models. People knew what they were doing. 

Worst model - portuguese building model this was only a point cloud

Latvian monument - the flowers show up only in the texture, not in the shape - because flowers are really difficult, but they knew what they were doing and it is very precise. 

Slovenian library - too MANY textures (500 textures, where 4 or 5 were enough). You should know how your equipment works. The raw data could be very good, but a person without experience worked on this. 

Do you get 3D models somewhere else? - Yes, if it is not that important, such as cars/trees/fences, but the important we scan ourselves. For buying, but mostly just downloading, we use Unreal Store  or Megascans. We ask people to give us credits, but otherwise no use of licenses. 

We have our own 3D viewer. You pay an annual fee and then you have your own domain. There is item management, you can make different view, also heavier or lighter models or GDPR safe. Adding metadata, annotations, export csv. Multiple models in one view. You can choose item or collection view. Storage is included with the viewer. 

Publishing on Steam creates a lot of traffic (f.e. TwinIt! Virtual Exhibition has over 100.000 views/downloads). People just stumble upon it. 

When downloading large models you could give a size warning up front. More than 200 gigabyte is for example to big (was exception with us too).