Validate Facebook import using StreetComplete or other means

We currently have 715153 ways originating from the Facebook import which still carry the “import=yes” tag.
As for these roads the classification was mostly guessed, we should verify it on the ground.

Could we think of a quest in StreetComplete which might help here?

Should we remove the import tag on all roads which have a reference added and assume that this implies on the ground survey? That would be 388 ways.

not really, as assigning road class requires wider context and some OSM knowledge

1 Like

A couple of issues I see with this topic:

A. There is no official way to indicate that a road segment’s classification has been verified on the ground.

For some of these old Facebook imports, I do sometimes update their classification and try to remove the import tag, but it won’t prevent someone else to change again the classification remotely and possibly with an incorrect one (like Grab did for thousands of them)

For my OsmAnd plugin, I use the vague source=GPS to indicate that a road segment has been confirmed by a GPS trace, but it does not mean the classification will be correct. But I can reproduce the source tag with public GPS traces.

B. As far as I remember, StreetComplete is great for adding these extra useful tags (lanes, surface) but I believe there is no way currently to pick a classification because rules would differ greatly between countries.

If we can make very detailed and reproducible steps to help StreetComplete users identify correct classifications in a new quest implementation, it would be fantastic.

However, we clearly don’t have these steps yet in our wiki, and at the moment you will not find 3 regular mappers in the same area picking the same classifications, so we are still very early.

Your reply sounds valid. Even among experienced mappers we have sometimes issues in finding the right category.

I still think that this tagging might help us to potentially identify badly tagged roads which should be surveyed on the ground.

One of my concerns is that a road is tagged in a class too high. Then routing engines mislead people on such roads. Getting stuck on such roads is certainly a killer for OSM reputation as reliable data source.

For high-ranked roads this doesn’t matter much. Whether something is a primary, secondary or tertiary usually doesn’t matter. Most of these roads should be fine for routing. So exact classification does not matter that much in this context of quality control.

A quick search turned up 16 highway=primary candidates. I might just manually check them against connected segments and remove the import tag. (overpass turbo)

Could it help to specifically guide the local survey on checking other road attributes like surface/width/lanes? That is probably less problematic than road classification.

I am still not certain what the condition should apply to justify the removal of the import=yes on these roads.

Could a StreetComplete edit giving the lanes/surface of a road be good enough to remove the import=yes? We won’t be able to find then Facebook imports that simple any more. But there might be other badly tagged roads as well, so this might be less of an issue at all?

StreetComplete asks about lane surface and also (with AR support or with mapper using tape and enabling quest manually) will also tag width of road.

I am aware of this existing quest. The AR measuring method was not self-explanatory and is maybe not that accurate. What are the preconditions for these quests to come up?

Would it be possible for StreetComplete to remove in such cases the import=yes tag?

probably no, as problem can be also with highway value

That would be something we would tolerate. There had also been massive directed editing campaigns by other actors besides Facebook. Look a bit in the old forum regarding complaints here. Grab/Globallogic was one example.

At least with some-one adding a surface tag, it is clear that road exists. And review of track vs unclassified is a larger topic on its own…

  • mobile phone must have ARcore support (or quest must be enabled in Settings / Quest selection and display order, but then it needs manual measuring as AR is not present)
  • OSM ways must match specific tags for quest to pop up

Would it be possible for StreetComplete to remove in such cases the import=yes tag?

Probably not as noted before (but also because that tag might have been used elsewhere in different situations). What you could do, is to use for example overpass-turbo see if the object was surveyed later (e.g. by existence of additional tags in this example) and then semi-automatically (after checking that it is ok) remove import=yes tag from those ways (e.g. using JOSM Download from Overpass API)

It is a larger topic, but it is probably good idea to take it into consideration now, otherwise you might end up with bunch of wrongly classified roads for which you no longer know (as import=yes was removed) that they are misclassified. (i.e. make the situation worse than it was before).

As for the possible solutions:

If you yourself are advanced user, you may want to try StreetComplete ExpertEdition fork (you can have it installed alongside the regular StreetComplete) which allows for custom overlays, so you can setup it to highlight for you all highway and import=yes and then on the ground select Show/Edit tags to manually update highway classification as well as remove import=yes for each of them.
But I would definitely not recommend that approach to anyone unless they have firm grasp of OSM tags in general and agreed highway=* classification rules for your country in particular.


The only way to verify if a road exists is to match it to public GPS traces. From my direct experience mapping in northern Thailand, most old import=yes are actual roads and not a concern, so I think it’s ok to just remove the tag. As you said, classifications are often wrong but that’s another topic.

I would not remove the tag, just to remove it. There is information attached to it that the feature came in via Facebook remote edit

As Facebook data was said to have lower quality, I wonder how to improve it. As we can identify the data: what quest in SC could improve it to a level we no longer need the import tag?

Target audience would be a mapper with little experience but wanting to help improve OSM. So it should be doable without expert knowledge.
This is how I came to the idea about physical road properties.

Other ideas what kind of quest might help?

1 Like