Sketchup Blog - News and Notes from the Sketchup folks

Updated Acceptance Criteria for 3D buildings in Google Earth

It was over five years ago when we came up with the initial Acceptance Criteria for photo-textured 3D buildings in Google Earth. Since then, we’ve learned many things and have also made many improvements to the 3D modeling process—including the release of Building Maker and two updated versions of SketchUp. Given all of these changes, we realized that our Acceptance Criteria were due for an overhaul.

Our new Acceptance Criteria have been completely rewritten with the goal of making them clearer and easier to follow. Issues relating to photo textures, permanence of structures, splitting, and entourage have proven to be the most common areas of confusion in the past:

Photo Textures

Our new minimum requirement for photo textures is more lenient than it’s been in the past. Photo textures are only required on upward facing surfaces of the model and on the main facade. We encourage you to photo texture the entire model, but we also understand that it may be difficult to get accurate imagery on every side of some buildings. Any remaining, non-photo-textured faces in your model should be painted with realistic-looking textures that match the color and look of the building in real life.

Permanent Structures

Beginning today, only permanent structures will be accepted. As we constantly refresh our satellite imagery, temporarily-positioned entities like vehicles and people don’t belong in Google Earth’s 3D Buildings layer.

Model Splitting


From now on, all submitted models should contain only one structure each. Each structure should be uploaded as a separate model file. This includes properties that have multiple buildings on them such as a house and a shed or garage. If buildings are all connected in a city block, they should be split into separate models based on building type, function or address. When our review team is assessing connected block models for splitting issues, we will look at the facade and roof textures to see if there are changes in material that signify where a split should have occurred.

Entourage


In addition to splitting buildings, we are now requiring all models of trees and other permanent entourage (such as signs, light posts and benches) to be uploaded separately from the buildings with which they may be associated. This ensures that when another building is uploaded in the same location, we are only judging the quality of the new building model versus the original. It’s a shame to have to remove good tree models just because they’re attached to a building model when a better building model is submitted that doesn’t contain trees.

Also, multiple, related trees and other entourage objects can be grouped into a single model as long as they are located in a relatively concentrated area. This means a single model can contain all the trees for a block or a park, but it shouldn’t contain all the trees for multiple blocks or an entire city. Remember that only permanent entourage is acceptable—cars and pedestrians move around, and thus don’t belong in Google Earth’s 3D Buildings layer.

Other improvements

One other big improvement we’ve made is the addition of tips and suggestions to each of the thirteen individual Acceptance Criteria. If a model you submit isn’t accepted, you’ll receive an email notification (opt into these emails via your preferences) that includes a direct link to concrete information about how you can improve it before you re-submit.

What about models that have already been accepted?

To help make this transition easier, we won’t be going through all the models we’ve already accepted in order to remove ones that fail to meet the new Acceptance Criteria. If your model has already been accepted, it will stay in the 3D Buildings layer until and unless it is sent through the evaluation process again. There are four actions which can cause a model to re-enter this process:

  1. You make an edit to your model and re-upload it to the 3D Warehouse, replacing the previous version.
  2. Someone else submits a model in the same location as your model.
  3. Periodic terrain and aerial imagery updates cause your model to go through our automated alignment process.
  4. Someone clicks the “Report a problem” link for your model in Google Earth.

It’s still a bit of a subjective process

Keep in mind that judging 3D models is still a difficult task and is prone to subjectivity. All submitted models are reviewed by real human beings who take time to ensure that they meet our standards. Because human beings sometimes make mistakes, we have a way for you to appeal negative judgements. If, after reviewing the Acceptance Criteria, you feel we’ve made the wrong decision, use the “Tell us why we’re wrong” link (at the bottom of the model’s 3D Warehouse page) to ask us to take another look. You’re encouraged to include links to photos of the actual building or other online resources to will help us to understand your point of view.

We know how much time and love goes into making beautiful 3D models for Google Earth, and we greatly appreciate all the effort you put into your work. Here’s hoping that the changes we’ve made will make for a smoother, more enjoyable geo-modeling process for everyone.

Posted by Jordan Van Wyk, 3D Modeling Specialist

Permalink | Links to this post |
The comments you read here belong only to the person who posted them. We do, however, reserve the right to remove off-topic comments.

19 comments :

Thomas Costick said...

I appreciate the effort that has gone into these revised acceptance criteria. I did my first geo-model a few years ago and have been intermittent since then and, returning to modelling, I'd been looking for the latest definitive criteria, particularly since I'd seen accepted models that included trees.

One thing not mentioned here, though: terrain. What would be the right approach to take when several buildings exist in close proximity, with a shared terrain adjustment? Should the terrain be submitted separately from the buildings?

sebatectura said...

well, I have the same question that @thomas costick.....

Jordan Van Wyk said...

@Thomas Costick and @sebatectura

Terrain should be avoided if at all possible. With that in mind, I do understand that the terrain data in Google Earth is sometimes far from accurate.

To answer your question, no, you should not submit terrain models separately from building models. It is acceptable to cut the terrain into sections that correspond with each separate building model. When all the models are uploaded they should fit together like a puzzle. As stated in the new criteria, a terrain model submitted without an associated building model is not acceptable.

Furthermore, to avoid a long review process, you should never submit two separate models that overlap with one another. All models undergo a duplicate check. If there is a terrain model which overlaps with multiple building models, it will result in increased review time.

SnowTiger said...

I'm happy to see a new Acceptance Criteria that takes into account current technologies etc. It's probably better if all reviewers are all working from a new but same perspective.
However, I must say that I disagree with the decision that properties (especially residential properties) that have multiple buildings, must now have each building uploaded separately. That doesn't make any sense to me at all. It would certainly affect several of my own models, including several recent models.
It might even effect whether or not I get contracted to model for a client who might prefer to have all their respective property structures shown in a single model. At least they reside on a single real property.
Maybe it's just me, but I suspect there will be more grumblings about this specific new rule.
That said, I hope Google is open to the idea of revising these rules (or at least that one "IF" others also disagree.
SnowTiger

Thomas Costick said...

I agree with @SnowTiger re the separation of buildings into individual models. It doen't make sense when structures are contained within a distinct area of real estate, e.g. domestic and commercial premises.

I'm currently working on a model of a scientific facility with multiple structures and significant (very necessary) terrain adjustment. In the past I would have submitted this as one model. Some structures are identical and share a common definition, and it makes no sense to submit these individually.

We can only hope that the reviewers at Google will be pragmatic, allowing exceptions where this makes sense.

oink said...

@Snowtiger, It makes sense because if one of those buildings is demolished and replaced then a new model can be made and inserted just for that property. Doing a single "block" consisting of several different buildings offers no flexibility in this regard.

This shouldn't affect your own contracted work, because you can split your single work into its constituent parts for the purposes of submitting to Google. It's a little more hassle, but that's about it really.

shakar12 said...

Hello,
We facilitate the provision of independent analysis to support expert testimony, regulatory or legislative engagements. Frequently, this work includes economic, financial and statistical studies of varying data analysis, technical and http://www.palmerleasing.com.

StrayKatStudio said...

Good criteria, Google. But would you accept airplanes that are part of an airplane museum? Those old things don't really move.

Also, boats. Some boats have had the same docking slip for 10 years and their business relies on customers being able to find them on the labyrinthine docks for their day trips.
Would there really be a downside to hosting boats with steady docking spaces? It looks awesome. It's akin to trailers in trailer parks. Technically, they are "mobile homes", but realistically, they're pretty much going to stay there. So could a modeler pull off submitting a boat, if there was corresponding documentation showing that, like a mobile home, it was a reliable geo-referenced structure? I'd bet the boat I'm thinking of will stay the same a lot longer than some of the trees that get hosted! =)
What do you say? Is this a loophole? Because I know a modeler who is a master of curved photo-textures, and GE would only get cooler by letting him do some boats and planes.

Jordan Van Wyk said...

@StrayKatStudio

Good question. It is only temporary structures that we no longer accept. If a boat or airplane is permanently docked or on display it will be accepted.

googlenaut said...
This comment has been removed by the author.
web30167 said...
This comment has been removed by the author.
Klip said...

when a block of buildings is split in separate models
the sidewalls will overlap, causing Z-fighting.
How do I solve that ?

Jordan Van Wyk said...

@Klip

You are correct that z-fighting would occur on the adjoining walls. This is not a concern because the z-fighting will be hidden by the adjoining building and not visible when viewing the model in Google Earth. To answer the underlying question, models will not be denied acceptance due to z-fighting in this circumstance.

For an example, please see the models at this location: http://goo.gl/JH9Ak (this link requires Google Earth Plug-in to be installed. To view in Google Earth, enter these coordinates in the search bar: 42.35053,-71.098787)

Generalfeldmarschall said...

What about skyways? Like the two bridges between the Hofkirche, the Royal Castle and the Taschenbergpalais in Dresden, for example.

Should I upload the skyway as a separate (floating) model, or with one building even if it looks as a floating appendix?

Jordan Van Wyk said...

@Generalfeldmarschall

For buildings with connecting skyways, the two buildings should be split as separate models but it is acceptable to connect the skyway to one of the models.

Contrarian said...

Hope you're still reading this, Jordan.

I have over a dozen accepted models which consist of multiple buildings (I always modeled entire blocks or half-blocks, if there is an alley).

I'd like to reposition some of those models to match the current imagery. Do these new criteria mean they'll all be rejected if I move them?

Splitting them all would take many hours, and if I had to do that, I'll just leave them where they are.

Thanks!

Jordan Van Wyk said...

@Contrarian

To answer your question, yes.

In order to be fair to all of our 3D modelers we must judge all models to the same standard going forward. If you update your models they will go back through the judgment process and be judged according to the new criteria.

I understand this can be discouraging. However, before you decide to not edit your models you might want to try splitting a few to see that it really can go quite fast. There are some helpful tips at the bottom of the splitting criteria page: http://support.google.com/sketchup/bin/answer.py?hl=en&answer=1267399

If you do decide against editing your models to avoid having to split them I would suggest that you make your models publicly-editable. You might even consider posting to the 3D Warehouse forum to find others to help you update your models. You can find directions on sharing your models here: http://support.google.com/sketchup/bin/answer.py?hl=en&answer=105700

cheers,

Contrarian said...

Thanks for the response, Jordan.

Ok, I'm gonna make a pitch to encourage you (Google) to re-visit that "no multiple buildings" policy. So if your minds are firmly
made up on that issue you can stop reading here. :-)

As I understand it, the chief rationale for the policy is to enable the model reviewers to make apples-to-apples comparisons of competing models of the same building. I've also heard that Google wants the models to serve as "containers of information" related to the building and its tenants.

Those concerns certainly seem reasonable on their face, but what are you losing in exchange?

Here are some disadvantages of requiring that buildings which are parts of a cohesive group (parts of the same project or property, connected by common walls, etc.) should be modeled and submitted separately:

* It is inefficient. Common wall buildings will have hidden geometry, increasing total file size and rendering time.

* Duplicate textures. Often buildings which are components of a single project, such as a college campus, industrial park, apartment complex, etc, will have the same textures on several of the buildings. If the complex is uploaded as a single model, that texture need only be uploaded (and downloaded by GE) once.

* Inaccurate alignment and poor integration. Common-wall buildings along a downtown block are unlikely to be properly aligned and scaled if they are all modeled by different modelers using different data (or, more likely, guesses) as to dimensions. E.g., awnings, parapets, crowns, rooflines will not meet at the proper places. Nor will shadows and color balance agree if the texture photos are shot by different people at different times. Result: a row of incongruous models lacking cohesion and thus realism.

* Terrain mismatches. Some buildings require some terrain "repairs." (That applies to all downtown buildings in my city, Spokane, WA, situated along the Spokane River. GE's terrain contour is grossly inaccurate there, and unless it is corrected, those buildings will either be floating 40 feet in the air on the river side, or buried up to the 3rd floor on the street side). When this occurs it always affects all adjoining buildings along a stretch of the river. If each of them is modeled separately, by different people, those terrain fixes will disagree and probably z-fight.

Of course, if the models in a group are all made by the same person as one model and then split, these last two issues would not arise. But then, neither would the "apples to apples" issue.

* "Holes." Modelers are drawn to buildings which are architecturally notable or are local landmarks. The inconspicuous, 3-4 story structures which make up the majority of any street scene, which together generate its texture and "life," will be ignored as subjects for separate models. So instead of cityscapes you'll get terrain imagery dotted with models of a few notable building and otherwise barren. Zoom into Des Moines, IA, for an example of this (take a look at Spokane, too, in street view. I think it is one of the best *hand-modeled* downtown areas on Google Earth).

For most urban buidings --- I'd guess 99% --- there will be no competing models. Indeed, most of them will never be modeled at all unless part of a larger project.

(cont'd)

Contrarian said...

* If Google decides to remove previously approved models which comprise multiple buildings, just because they've been moved to

match current terrain, it will lose *most* of the models in the best hand-modeled locations on the globe. Most of those buildings

will never be replaced, until Google decides to auto-model those cities.

I'm presently working on a model of Sacred Heart Medical Center here in Spokane. It comprises something like 15 buildings, many

with common walls and others connected by skywalks. The entire complex is situated on a steep hillside, with the buildings all at

different elevations. Hence the first order of business has been creating a terrain overlay which will allow all those buildings

to intersect at the correct (relative) elevation. It would be absurd to try to construct all those terrain segments separately, or

to expect models created by independent modelers to properly integrate.

I should think you'd want to encourage, rather than discourage, modeling complete groups of clearly related structures, rather

than inviting submissions of separate elements of such complexes by different people, and thus sacrificing their functional and

aesthetic unity. You said above that even garages and sheds on a property should be submitted separately. Who is gonna model a

garage on a residential property but the person who modeled the house?

My suggestion: re-define "building" to mean a "logical building," rather than a physical building. I.e., a structure or group of

structures which are physically connected, functionally related, or are components in a larger project, such as the dozen or so

buildings in a garden apartment complex, or the connected buildings in a shopping mall. Then encourage modelers to tackle the

entire "logical building."

Just a thought!

I agree that splitting a new model before submitting is no great trouble. But re-visiting all my existing approved models, of

which there are about 20 comprising probably 100 buildings, is a real downer. (Look for models in the warehouse in Spokane by

"GaryM").

Thanks for reading.