The Huge, Unseen Operation Behind the Accuracy of Google Maps
wired.com
wired.com
There are projects based on OSM for routing, geo search, map editing, etc. that are also free. Definitely a lot of room for both programmers and non-programmers to contribute.
I know for a fact this isn't true. What was your source for this?
As an engineer-developer hybrid, I found OSM mightily helpful when pulling midwest town data and running tornado strike simulations. I encourage contributors to look beyond the roads -- other infrastructure such as power lines or towers are rare to find in the data but can be extremely useful.
http://wiki.openstreetmap.org/wiki/Humanitarian_OSM_Team
Quoting from the wiki page:
"From the beginning of OpenStreetMap it was anticipated that open, free map data would be a tremendous benefit for humanitarian aid and economic development. The idea was proved during the Haiti earthquake in 2010. HOT was incorporated in the immediate aftermath, August 2010, as a U.S. nonprofit and became a registered 501(c)3 charitable organization in 2013. Anyone is welcome to contribute to the HOT mission via our Tasking Manager - all you need is an OpenStreetMap Username. We ask only that you try to uphold the same code of conduct as our members, see our HOT Membership Code."
Surprisingly for me, quite many important buildings can be discovered without local knowledge (i.e. without "being there physically"), purely based on aerial imagery and a specially crafted guide. See e.g. the following link, with some hints and clues on how to recognize a school for a current (urgent!) pre-typhoon task http://tasks.hotosm.org/project/804:
http://wiki.openstreetmap.org/wiki/Humanitarian_OSM_Team/Typ...
There are public datasets and imports that happen sometimes, but these are done by people who have experience in it, and usually in cooperation with the local government GIS that controls the dataset.
http://www.openstreetmap.org/?mlat=59.4867&mlon=17.8986#map=...
Read more: http://wiki.openstreetmap.org/wiki/Import
People are happy if you manually enter them.
We have looked at some alternatives such as geocod.io, but quality seems a bit lacking and they don't offer some features we need (but which we would surely be able to develop ourselves if we could use OSM as a primary database).
The advantage of OSM, is that if it's not returning the right data, you can fix it, and import it into your own hosted version.
I'm one of the Geocod.io founders. I'm very interested in hearing about the features you're looking for and discuss data quality. Feel free to reach out here or at hello@geocod.io
While not as detailed, for contrast I want to post how Nokia's HERE maps builds their maps.
http://360.here.com/2014/11/17/made-usa-people-fargo-make-ma...
That is from the company blog. It shows a couple of our tools, but not nearly all. I work for HERE maps and my opinion is my own.
I just thought that people might be interested in how other companies like HERE build maps. Despite all the automation (at both google and HERE), it is still a human intensive process and there are people on the other end making it as correct as can be. Also, I can't go beyond what is written on the public blog as I am not a company spokesperson.
That being said, you'd be surprised how little you need to refer to your own company by name internally, especially in an ambiguous manner. Usually its just "Hey, can you put this up on the site?" or "It would be cool if the product did X."
That all being said, while Google Maps as a data source is amazing, Google Maps as an app and or web-site has a lot to be desired. Up until a few versions ago they had a way to add "My Places" but no way to actually access those places from the app (what?!). They have no search history (still, today), they do have suggestions but they are just somewhat useful, they have no compass (got removed), plus they went all "minimalist" and hide all the UI elements (and removed things like offline maps, for nearly a year).
I've actually been forced to stop using Google Maps as GPS as the app keeps closing during navigation and when re-opened it has "forgotten" what I was navigating to or previously searched for. So now I am forced, mid-journey to exit the freeway just to re-open Google Maps, re-search for my destination, and then re-start navigation just hoping it won't re-close-randomly before I get there...
It doesn't crash. It just closes. It is just gone. Like "poof." Then of course the phone goes into standby as nothing is keeping the screen on. At least if it had search history the situation MIGHT be recoverable, but nope...
An odd thing is that if I've searched for something on Google Maps on the laptop and then open their phone app I only have to type one letter and then the name pops up of what I'd searched so there's evidently some sort of history, just not very well implemented. Another funny glitch I had with driving directions is it quite often says 'turn left' about 4 seconds after you have passed the turning which is not quite as it should be. I sometimes switch to Apple Maps of all things though I don't think their traffic info is as good.
1.) Send directions to a mobile device. There are many times where I search for directions or map out an alternate route on my desktop. The only way I can access those directions is to email it to myself then open the email on the mobile device. Why not just store the data and let the user retrieve it on their mobile device?
2.) Better accuracy of property in rural/suburban areas. It's odd to say this but Bing Maps is actually much more accurate in rural areas.
3.) More fine grained control over street view placement.
4.) Offline maps! Not everywhere has good cell coverage.
Agreed on the apparent deprecation of social/offline features, but the navigation does seem to be getting better.
However, your starred items do remain even while offline. So for travelling without data, given some planning time, you can star a bunch of stuff while you have wifi, and then navigate using the offline map without search.
I've tried many other solutions, and this really was the simplest and best for me.
Yeah. How is this not a thing?! I imagine a lot of people search for things on their desktop and then want to "push" it to their phone for navigation, but yet I have to re-enter the zip into the phone to do so.
Firefox: www.foxtophone.com
Chrome: https://chrome.google.com/webstore/detail/google-chrome-to-p...
I'm not sure how the could integrate it more (or even less) with Maps or Android or what the phone being off or locked has got to do with it. It syncs the same way all Google services do. I suppose it may be working less well on an older or oddly (manufacturer) customized device.
I think the point of this is to be reassuring: "Hmm... I wonder if it would be faster if I took a right here to get around this backup? Ah, Google says no."
That's increasingly important as the live traffic data gets better and it's worthwhile to use navigation to pick one of several routes you already know.
Google is still not perfect in this regard, but I'm certain my batting average at being able to "beat" its recommended route to a frequent destination has been trending down for some time.
Maybe the good guys at Google need to learn about caching.
But, sometimes I just go to Bing because I know it will load fast and I can scroll around super fast, too. Scrolling around in Google maps is always slow. It's like they are loading way too much stuff. All the 3D buildings...useless.
I've had a similar issue occur several times in the past month. The app doesn't close, it just stops providing navigation instructions. The worst part is that it fails silently, and it seems to be related to poor cell service. It was a little scary when I realized what had happened, as I was driving on icy roads in an unfamiliar area at night. A notification that navigation has stopped would have saved me time and headache.
I'm also not a fan of how it changes your route without telling you. I intentionally selected a routed that took major highways, but after stopping for gas the app changed my route to a shorter (by distance) trip that was on small, poorly maintained roads through rural areas (with very poor cell coverage). When combined with the navigation silently failing in poor cell coverage, this has made for a very poor user experience.
You do need to turn it on originally, but I believe that option is also fairly prominent...
I might jump at the chance. My stress levels are correlated much more strongly to traffic than trip time.
At least on iOS, there is. It even syncs with your history on the desktop maps.google.com so if you looked something up on your computer, it's right there.
Just for context, the first image with the red and green dots--- each dot represents a 360-degree "street view" image. While consumers only get half-a-dozen a block, the operators got access to all of them. (Green are HD, red are normal def.) It was really useful for verifying that businesses had closed because the streetview dots also had the exact time that they were taken.
There was also deeplinking in Atlas to a specific dot and direction vector. So when I was submitting a change to a business, I could link to the exact sign that I saw to prove that "Mama's Teriyaki is on _that_ corner", which helped whoever QA'd my change.
It was a great first job out of college given that my degree wasn't very relevant to anything. (A lot of my coworkers in the same spot--- young, a degree that wasn't too useful, trying to find anything to do.) Out of the 400 contractors in that first Kirkland "class", I see tons of them around the Seattle tech industry now. It's been a surprisingly good network to have.
At the end of the day it was still data-entry / data-validation, and mind-numbingly boring.
I don't quite remember the numbers, but it was something like 40 a day? The slower and new peeps would be around 25, and the fast people could do like 80 a day. But there weren't any productivity bonuses besides not being let go, so me and my buddies would do about 10% more than the average and then goof off on the internet or do homework for our python night class.
The Local team used Atlas, as well as the GroundTruth road-placement team, and the QA teams. There were different map layers and controls you could bring up.
Business category was funny because it had the most subjectivity to it. There was also a team in Hyderabad (that wouldn't make phone calls), and I once got into a dispute about whether "Olive Garden" is an American restaurant or an italian restaurant. :D
Yup, so part of the Atlas history for a POI is seeing problems that people have submitted onto a listing. When I was doing this a few years ago we had a humongous backlog of user feedback, but I'm under the impression now that they're mostly up to date.
Spammers were a huge problem. For businesses that don't want to advertise their brick-and-motor (a locksmith comes to you, not you to a locksmith) the policy was that they'd get one POI at the center of the city or zipcode they operated in. You'd think that locksmiths are bad, but florists are worse. Florists! They sell flowers to people! Horrible spammers. They also like to make fraudulent listings--- so they'll make a listing for "Competitor's Flowers" but have the phone number go to their business "Evil Flowers LLC".
Regarding florists, they're bad, completely agree. I weeded out a few mega-spam networks a few years ago before I gave up and focused on locksmith spammers. Garage door suppliers, handymen, carpet cleaners, movers, escorts, and lawyers! Lawyers are pretty bad, especially the ambulance chasers/accident attorneys.
Right now you can use Google Maps on your phone to add a missing place. This place will be approved (usually within an hour) by an Anonymous account bearing the "G" logo. Now while the POI may not display unless you are actively searching for the place or it's specific category, it will still appear on the map. This makes it even easier to crowd the map with useless information. I am sure it's an insult to those who painstakingly edit the map in MM with accurate information, only to be denied be an LE who knows nothing, and then seeing the useless entries like people's houses and spam get added immediately. What is Google's reasoning behind this?
It's all fairly easy to spam Google Maps from other avenues. I've seen bulk uploads of 100's of spam listings, the Google My Business product is also easily gamed, and the LEs don't appear to closely scrutinize locations in order to ensure that the business is where it's claimed to be, even though they have access to more detailed street view imagery and presumably can just Google it. Google in general, and LEs in particular, seem to have a policy of preserving all data, no matter how bad/spammy, and approving new POIs, even if the POI is of a questionable nature and isn't supported by the facts, which, again, points to a deficit in their training in being able to identify spam features for Maps. Once you know what spam is, it's fairly easy to identify, and the actual QC checks, primarily using government licensing databases and other resources, takes about 30 seconds or less to verify listings. The hard part, at least on my end, is removing it. Google loves to hang on to spam.
(Though it makes me feel better about taking an angle grinder to my door when I lost my keys a few years ago.)
(Incidentally, you made a wise choice.)
Similarly, local florists are great, but many of the florist ops on Google Maps are just resellers running out of a call center who tack on an extra fee for ordering the flowers for you.
Basically any business that is aimed at homeowners, but tends to be located away from residential areas, has a big inventive to create false locations.
You survived the place hope goes to die? Congrats! It sounds like an unholy amalgam of data entry and a call center.
Imagine a world where all the effort Google, Apple, Microsoft and Nokia put into their maps all went into OpenStreetMap instead.
You are right because there would be a huge boon. This could be said for many other things too, like drug research for instance.
Rather than pining over a wish, we should come up with a better open source/crowd sourcing method if we really want them.
Here you go, one method for crowd-sourcing map data:
www.openstreetmap.org
Maybe we can't! Capitalism is pretty damn good at some problems, optimal even. But it's not flawless, and it is worth noticing and thinking about places where gains appear to be possible.
The way you get the incentives correctly aligned is to make sure that the people making the decision (i.e., the company building the service) gets reward commensurate with the social utility, such that it is in their rational self-interest to do the socially-useful thing.
At which point, they are no longer doing it without compensation.
Many GPSes have "live google" functionality. I have one that allowed me to search for the nearest gas station or restaurant while on the road. There is huge value to Google from that.
That's where you got it all wrong; the same common sense argument could be used against using (and contributing to) open source software in companies, and yet the most successful open source softwares are supported by companies. The thing is, contrary to the common sense argument, there is value in not reinventing what others are already doing. That's also a great way to find technical talents who have first-hand knowledge of the tools you use.
OSM is very similar here: the data is openly reusable by anybody and any change benefits everyone.
There can be a good reason for not contributing and using OSM, and that would be because Maps is such a central product to Google that they don't want anything shared with anybody else (plus they already have a lot of data and are considered as the best in class, rightfully or not)
OSM is an interesting case. I would have thought that the sheer effort required to keep map data up to date across the world would make it severely impractical to have open map data. They seem to be doing a decent job plugging away, but I don't have any data on how good they are at getting really accurate data and keeping up to date with ever-changing road conditions.
Capitalism is doing much better than the other existing systems at solving the massive coordination problems of a large modern society. However, we can also tell that it is imperfect and involves some clear examples of wasted effort. Noticing this waste is valuable, even if no one has a good solution at the moment to improve on the existing system.
What's the alternative? Should we just assume that our current economic system has an optimal set of incentives that we could not improve upon?
Its been said that Maps is likely Google's most expensive product to maintain. Just think about the scope of driving every road in the world as often as possible.
It will only be replaced by OSM if OSM gets to be better than Maps on its own.
This isn't unique to Google Maps; the same goes for competing pharmaceutical companies, educational institutions, and everything else. Imagine if they were all cooperating rather than competing.
However there are also some benefits to competition. Notably, the impetus to take risks of private capital in order to build new things / try new ventures. Also, the presence of choice in the market (in this case, you can choose which maps you consider more reliable).
I'd love to see more exploration into economic systems involving cooperation and choice, instead of competition. Can we have independent organizations motivated to produce different-but-similar goods, while simultaneously sharing their efforts?
The whole problem is the incentive to innovate diminishes greatly for non-profit, single sources for most products. Thus, we have capitalism and "different commercial vendor silos" for most things, including maps.
Waving hands wildly: imagine a mapping database that was owned by the government, free to use for non-profit uses, and licenseable by corporate entities. Here's the twist though, the license fees are proportional to the government doing a certain amount of upkeep/improvement on the database, and you are free to provide the same work block instead of a fee. That way, if you can do it cheaper, it will save you money up front and if your competitor can't do that, it's a competitive advantage.
I haven't thought about this model carefully just pointing out it isn't necessarily as black and white as you suggest...
It does happen at schools, say, where the players for each team in P.E. class are picked at random, but fans get engaged with the team identity and the individual players, and team abilities depend in part on long-term shared practice.
It's hard to believe that the space programs of the 1960s would have been better had the Soviet and U.S. programs joined forces for a single space program; also, obviously neither space program was based on the direct market forces of capitalism.
The military is also not run under a capitalistic basis, and the concept of "single source" for a NATO or UN force is meaningless. It's even a stretch to say that a combined US Army/Navy/Marine operation is a "single source" given the inter-service rivalries. I'm not convinced that a single world government with its own military would be better.
1. That Broad St no longer exists for a significant portion of its length.
2. That Mercer St has been two-way for months now.
This is the major east-west road in that area, and they can't even get that right.
Yes, I've reported it regularly for months now.
TomTom seems to be the winner with realtime traffic data accuracy, besting Google, and Google beats Apple, although not by much.
Google has lead in POI info, but that lead has diminished to the point of irrelevancy in the last year. Apple and Navigon have the best visuals. Google has the most naturally conversational voice guidance. And while those are nice, I still tend to prefer TomTom's accurate traffic data.
In looking into getting it fixed I heard that the contract for getting the map data prevented Google from making any modifications to the map data. Apple Maps might be in a similar situation -- aware of the problem but with their hands tied. Map data has a lot of legal restrictions for reasons I don't really get.
If you want a bad case, ask for directions to the town of Langley, which is placed on the wrong island altogether.
Some folks have said that Ingress, the Android and now iOS, real world geo game. Was in part an exercise in data collecting. Get people to photograph land marks in the real world to make "portals" while collecting GPS data associated with public landmarks for maps. And then use that data to localize things in satellite imagery, etc.
Makes for an interesting data set.
I know business considerations are what they are and often independent from others, but nothing shook my belief in Apple's commitment to quality as a defining principle more than the decision to replace Google Maps with Apple's own half-baked beta offering. Does anyone think they'll ever give maps & cartography the kind of attention to detail that Google's given it?
Not so amazing. Just the data is amazing. The app sucks.
Which is exactly what they've been doing, all while hiding as much of the interface as they can and obscuring as many affordances as they can.
How Google Builds Its Maps—and What It Means for the Future of Everything
http://www.theatlantic.com/technology/archive/2012/09/how-go...
I think most interesting point is not covered in either of them: How much of this effort devoted towards human curation? The numbers are believed to be ~1100 full time employees and 6000 contract workers[1]. These are huge numbers compared to most competitors in the market. Assuming each correction to a map can take 15 mins on average, you can easily make 100,000 corrections a day. Again, assuming there are top 2000 cities in the world where most of the queries originates, this is about 100 corrections per city per day. This would guarantee Google maps best of the best freshness, precision and recall on most metrics. With about 10X-20X larger curation force plus algo engineering, likes of Apple or HERE have no chance. In a way this also shows Google's leadership wisdom. Maps are the most important thing on mobile and even on web. Most companies don't get this and provide minuscule budgets citing no potential revenues (for instance Ballmer cancelled Street View like effort at Microsoft). By the time they wake - if they wake up - it would be to only find that they have been outrun by such a huge gap that even a decade won't be enough to catch up.
[1] http://www.businessinsider.com/to-do-what-google-does-in-map...
Why would an AI and automation company be better equipped/able than a commodity hardware company to run a maps/routing service?
Is it really that crazy a notion that actually driving the roads and collecting all kinds of data from it yields higher accuracy maps than just licensing less robust data sets from others?
Sure, keep analysing that shop front.
I focus on street view because it is a big source of data (per the FA) and a testament to the comitmment and effort that's gone into it. In the U.S., how many other map makers actually drove 99% of the roads on thier map? Maybe that doesn't make a difference to you, but I think the central premise of this article and the general consensus of most users is that it does.
Investigators determined that web-based mapping software Expedia was used--based on symbols used to mark highways and such--ruling out Mapquest, Yahoo, and others.
A single person had clicked on that intersection in Expedia in the 5 days before the map was received by the newspaper. They traced the person presumably using the IP address.
This was back in 2001.
It's odd that we still have unsolved street crime given that everyone today carries a tracking device sending back real-time geolocation data.
Just wait for the self-driving car tests to ramp up. I think one of Google's early deployments will be a fleet of self-driving cars equipped to take streetview photos - this should enable something like weekly redrives of cities.
Additionally, Google is now providing wi-fi for Starbucks, so they really should have all the locations.
The hipster cafes, though, seem harder to find automatically.
I see your point, but the reality is that my users are there and these are also products I use every day. It's in my interest to have them work correctly.
Apparently they still haven't fixed the whole "we have to go directly away from the destination for a little while" problem.
I doubt this. Out of curiosity I looked at my family's farm and Google Maps is missing most of our outbuildings. More generally I'm sure that they come up short on backyard sheds, pool buildings, structures in dense forests, etc.
Actually last I tried Uber/Google Maps sent you at the directly opposite corner of Paris, to the Quai Branly Museum, probably because there is something related to the anthropologist there (?).
Both Apple Maps and more interestingly the crowd sourced OpenStreetMap know about the "promenade". I'm curious to see how long it will take Google to discover this new street.
Not too suprising really. OSM can be updated in a minute to add new features.
They're doing a good thing, but unless you can see that Google has driven the route and has published street view, you can't entirely trust them.
[1] 37.515696 -122.050873
Using the GPS drains battery and there are huge privacy concerns. It's not an easy thing to do right. But I think it can be done.
Yes. Google deduces traffic congestion by the speed of Android devices in cars.
>"We have reviewed the coordinate reference system used by Microsoft, Google, etc. and believe that it is technically flawed. We will not devalue the EPSG dataset by including such inappropriate geodesy and cartography."
http://blogs.esri.com/esri/arcgis/2010/03/05/measuring-dista...
The zoom level bug you are hung up on was due to the way web mercator was implemented in Osmarender. This is why people are answering your concerns about teh zooming by pointing out osmarender is no longer used.
So no, don't use web mercator for analysis. But only worry about inconsistent rendering across zoom levels if you are using Osmarender.
All projection systems have "inaccuracies". It has literally been mathematically proven ( http://en.wikipedia.org/wiki/Theorema_Egregium ). Never the less, GIS people can, and do, use many projections professionally.