Hacker News new | comments | show | ask | jobs | submit login
What OpenStreetMap can be (systemed.net)
192 points by Doctor_Fegg 80 days ago | hide | past | web | favorite | 42 comments



As to humanitarian use for OSM and how anyone can contribute by small tasks (I've tried it myself, was surprisingly exciting), enabling faster and more accurate response and travel to people in need, see:

https://wiki.openstreetmap.org/wiki/Humanitarian_OSM_Team

https://wiki.openstreetmap.org/wiki/OSM_Tasking_Manager

As to work in progress on detection of features and automatic conversion to maps, I've recently found the below website:

https://blog.mapillary.com/update/2016/07/15/mapillary-osm.h...


Facebook and Microsoft have both undertaken large scale ML projects contributing data to OSM. Facebook has a pilot in Thailand:

https://forum.osm.org/viewtopic.php?id=61826

and Microsoft released 124 million buildings for contribution to OSM:

https://blogs.bing.com/maps/2018-06/microsoft-releases-125-m...

There's also been some work by individuals:

https://github.com/jremillard/images-to-osm

A key to the satellite imagery projects is that the imagery be suitably licensed.


We recently overlaid the Microsoft buildings over Boulder Colorado aerial maps and in some places they were pretty good. In others, where there was just the right amount of tree cover, it was pretty bad. Houses cut up into 2 or 3 or 4 chunks, for example.

It's a great data set, but it'll probably need some curation.


For sure, it will need curation. But even if it were just information about where buildings probably existed it would be hugely useful. An example use case is improving building coverage in a rural area.


Sure! I wonder if the ML system could make use of information such as lot size and shape (from the county), building sqft, number of levels, information about outbuildings (from the MLS or property tax), and could get smarter. And if it can tell a tree from grass.

"This small lot has 3 200sqft buildings separated by a tree on it, but is supposed to have a 1200sqft ranch house... Maybe those 3 buildings define the bounding box of the house?"


> It's a great data set, but it'll probably need some curation.

Some sort of crowd sourced, geo database... ;)


There is also a good buildings dataset from DRCOG that we're (OSM Colorado) importing into OSM. Reach out to us, or come to our next meetup if you're interested.


I'm excited about this direction! I do minor contributions and poking around my area I'm surprised by how much of it is done by very few people. How useful it is yet how lacking it is. There's a neighboring town that is absurdly well documented, it seems a group of people got together over a month and mapped the place out meticulously, including drawing all buildings etc. This town would work very well for these vector maps, my town probably not so much.

I wonder how to get more people involved, and excited, about OSM? How do we make it the Wikipedia of digital maps?

And does Strava/Mapbox give back? When Strava identifies common bike routes, or perhaps trails, and when Mapbox determines that roads have changed is that given back to OSM in some capacity?


> I wonder how to get more people involved, and excited, about OSM? How do we make it the Wikipedia of digital maps?

This seems to be the general sentiment of the OSM community, without evidence that merely adding more people will yield better results.

I personally was turned off from contributing due to the emphasis on "community", even at the expense of map quality. I blogged about this 7 years ago.

As with Wikipedia, and perhaps much more so, the community can be very fragmented (both online and, obviously, geographically), and the portion holding the real power is usually very insular.

This essay https://www.jofreeman.com/joreen/tyranny.htm recently popped up on HN in reference to Valve and "flat" or non-hierarchical companies like Valve, often with a caveat to ignore its focus on the feminist movement. However, in the case of another all-volunteer project, it may apply better without that caveat.


Before their privacy apocalypse, Strava made their traces available for use in OSM. Mapbox actually had a project where they added information that Strava reported as missing:

https://github.com/mapbox/mapping/issues/113

Mapbox also uses other data to find things to improve.


    >  I'm surprised by how much of it is
    > done by very few people [...] How do
    > we make it the Wikipedia of digital maps?
The exact same contribution curve applies to Wikipedia and OpenStreetMap. Something like 5% of the users to 90% of the work.


hmm care to be more specific ?


Did I misunderstand, or was that not the entire thrust of the article?

That vector tiles enable the representation of a subset of the data that a particular group care about (e.g. playgrounds) in order to encourage and enable contribution from a wider and more diverse community?


Seems to me like that's more of a new defining feature of OSM, which may incidentally attract more contributors, but that in and of itself is not lowering the barrier to entry for prospective contributors.


What makes OpenStreetMap shine really is their desire to map every man-made and natural object.

I'm talking about things like trees, garbage bins and industrial buildings.

Commercial maps has areas where they care deeply, and areas where there's zero care. OpenSteetMap cares about everything. If you are a hobbyst (of practically anything), sooner or later you'll find better representation of your narrow interests on OSM than on Google or Apple maps.

Google has awesome maps of shopping centers. OSM has awesome maps of tracks in woods and parks. You've got to choose your life style here.


Google Map's lack of geographic features has been off putting for me for quite some time. They can list every last commercial entity under the sun, but can't be bothered to get the name of a river correct, name high points (which would be useful for navigation), canyons/ravines are generally missing. Items like campgrounds tend to direct to commercial enterprises rather than dispersed sites.

OSM has all of that, it's great.

I haven't had time to really dig into OSM. One item that confuses me is the community-based editing which lends me to not wanting to touch it out of fear I'll cross someone (like Wikipedia). I do a lot of personal mapping on Google Earth of routes, points of interest, etc., just dumping GPS data out after walks. I would like to do the same on OSM, and store it as my own private overlay but I don't see how.


There's no personal mapping features on osm.org.

You could try http://umap.openstreetmap.fr/en/

As far as crossing someone, it certainly happens that people get territorial and there are lots of relatively pointless objections, but many people welcome improvements in the areas they pay attention to. And then a difference with Wikipedia is that in OSM, if you go look at something and establish ground truth, that's the gold standard.

You can look at http://resultmaps.neis-one.org/osm-discussions to see what a lot of interactions end up looking like. Usually polite and straightforward.


> One item that confuses me is the community-based editing which lends me to not wanting to touch it out of fear I'll cross someone (like Wikipedia).

This is what drove me away 7 years ago (actually crossing someone, not just the fear). It's interesting to hear that this is still a relevant concern today.

> store it as my own private overlay but I don't see how.

I would expect this is possible with the right software, but it may mean actually keeping a synced, private copy of the entire OSM database, which could get cumbersome.

You might also check if there have been any projects that have succeeded in doing some kind of dynamic/living fork and what technology they use.


I've been adding to OSM maps for about 5 years, mostly in the UK, but also on Humanitarian OSM. I've had no problems with crossing people.

I think it differs from wikipedia in that I am usually adding to the map (mostly paths and buildings), rather then editing existing objects (except when they are out of date) - so conflicts are not likely to occur.

Possibly in regions where the map is already 'perfect' there might be conflicts, but you are less likely to be editing such areas, I expect.


Does this include items indoors? Such as pinball machines?


You can not necessarily create a point for a single pinball machine (yet), but the Key:Leisure allows you to have a rather fine level of details.

https://wiki.openstreetmap.org/wiki/Key:leisure

It mentions "amusement_park" to tag areas with

"A venue with pay-to-play games, such as video games, driving simulators, pinball machines, merchandisers, etc."


Sure.

If you can go look at it and it is relatively static, that's good enough.

So trees are much more accepted than say flowers.

So far only 1 has been mapped using the obvious tag:

https://overpass-turbo.eu/s/ABP


There are ways to map things indoors, like corridors, stairs, and shops/facilities inside large buildings (eg. train stations)

https://openlevelup.net/


OSM has awesome maps of tracks in woods and parks and shopping centres.


Of course there's also derivates of OSM that would be impossible to replace with Gmaps. For instance the topographics maps with detailed path characteristics for offline use for hiking and cycling provided by https://www.openandromaps.org/en


I recently went hiking in the pyrenees, and I had IGN maps (french topographic service), randonnee edition maps (commercial, based on multiple sources) and the OSM hiking map available offline on my phone though OSMand.

I found the app the most accurate source. Though the other maps were better about displaying where guarded and unguarded mountain huts were.


The bit that people seem to be turned off by is what I like to call the 'Stonehenge is not a henge' arguement.

I started trying to contribute to openstreetMAP, to be told that where I was going wrong is that I was thinking about openstreetMAP as a MAP. I am such a fool. It has become a community mantra, and it just sounds silly.


>Most OSM commentary focuses on unimportant minutiae (layers, for goodness’ sake, as if it’s still 2004)

What's wrong with having layers on a digital map? I can see the problem of not having them.


The cited context is about segmenting data for editing, where right now everything shares one context. For display and most others uses, the data is processed to create convenient layers.


It's not hard to render OpenStreetMap in such fashion that there are layers. I don't understand what kind of explicit support is needed here.


Keep watch for the live streaming[0] of the State of the Map 2018 event tomorrow (July 28-30). They already have interesting videos on their channel such as using OSM for traffic simulation.

[0]: https://www.youtube.com/channel/UCLqJsr_5PfdvDFbgv1qp2aQ


I think, if it were feasible, OSM could use a really sharp CEO. I think it has incredible potential, but it needs strong leadership to determine strategy for differentiation, product positioning, and marketing. Microsoft or Facebook could make an investment in this direction and things could really take off. Just a thought.


I often wonder what parts of the world are well-documented in OSM because my city is sorely lacking. I'm sitting reading this in a restaurant that has been here for decades. Not only is the restaurant not in the data, but the building itself isn't, which is something that I thought was pretty well covered.

I've added a major post office in the past. Not a new one either.

It's really hard to get enthusiastic about contributing to it when the data is just abysmal here. I hear about people adding hedges and just am baffled that their area is so well-covered that a hedge is a notable omission.

I know, I know, do it myself. Be the change you want to see. But man, without some sophisticated AR tech I don't know how Indianapolis can catch up.


What motivates you contribute to HN?

There is an old proverb of "A journey of a thousand miles begins with the first step". Almost all contributions make a positive step for improving the map for everybody. In the time it took to write the comment you probably could have used your local knowledge and aerial imagery to add some details to the map.

I remember when I started adding data to the map in 2006 and it was almost all blank. I was happy to cycle around my local area with a GPS and spend time adding data locally to me. I did it partly because I was interested in the technology but I also had faith that when I needed to use the map someone else would have added the data that was local to them. As it turns out, this faith was justified and the OSM maps of the UK are great.

For OSM to continue to make progress there is an assumptions that there is a subset of the general population that are willing to contribute for the greater good. I know that isn't always the case, the argument of: "why bother when we all use Google maps" is common. As it turns out, what Google displays on the map and what they allow you to do with it is fairly limited.

As Doctor_Fegg argues, OSM is only just beginning on the journey to becoming an essential resource for worldwide geographic information. Right now things are still too focussed on providing a limited set of raster tiles which typically throw away half the data when choosing what can be displayed.

Vectors tiles are an obvious next step for expanding the visualization of the existing data and provide the client with greater access to the underlying data which might not be shown on a raster tile.

Another important point of open data is that it allows multiple sources to be combined together. OSM has never tried to collect everything. For example, SRTM is a better source of altitude information. There are other websites which provide open aerial imagery or street-view style videos. Openness of the raw data is the fundamental differentiator of OSM versus any alternative that I am aware of.

Just like open source software is increasingly relied upon by businesses and the internet as a whole. I believe that open data will also be seen as being a fundamental building block which will never decrease in value or adoption.


You could start with this app:

https://github.com/westnordost/StreetComplete

It's super easy and once you run out of tasks, you can dig in deeper with this one https://wiki.openstreetmap.org/wiki/Vespucci or just through the browser.

Germany is quite well documented.


> But man, without some sophisticated AR tech I don't know how Indianapolis can catch up.

It really only requires a few people who want to stay busy doing something that doesn't require of thinking to finish their day. In the early days of OSM, I mapped hundreds of local hiking trails. I also aligned extremely inaccurate street vector lines to the underlying aerial images for tons of cities. It's a very satisfying activity once you're in the flow.

Similarly, some madman has walked all the streets in my city and annotated all houses with the correct house number.

One of the best parts of OSM is that it's gathering data that is largely static: street location, and names, and number don't really change over time. So you don't need a large number of volunteers to improve a local map in a meaningful way.


> It's really hard to get enthusiastic about contributing to it when the data is just abysmal here.

Be careful what you wish for. Once there's enough pre-existing data, or many contributors actively adding it, you'll have to worry about offending the "community".

Right now, you can be that community and add what you feel is important in the way and order you feel is important, at least locally.

Alternatively, you could try to ETL [1] the data from other sources, though free ones aren't always available, especially if something like building data hasn't been put into digital form by the government. Even then, you might be in for a fight to get it for a reasonable cost, even if freedom-of-information or public-records laws are technically on your side.

[1] certain computer skills are obviously very helpful, but expect to QA your own automation.


Time to start mapping yourself! You'll be able to put things on the map. Just keep mapping, persistence will pay off.



Maybe, it is time to create an OpenGPS? Please, let me know if you know how to build one using RPi or something similar.


Making an OpenSource GPS is technically challenging. For an idea of what is involved see: http://www.aholme.co.uk/GPS/Main.htm

Here are some other similar projects: http://sdrgps.blogspot.com/ https://github.com/hahnpv/gr-gps https://github.com/hahnpv/SoftGNSS

Personally I would rather buy a $10 modules off eBay or a Chinese website which provides the GPS data in a standard NMEA serial stream.


OMG, now I see why we are so dependent even for old technology, and I now understand the great success of Arduino.




Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact

Search: