Hacker News new | past | comments | ask | show | jobs | submit login
Apple's New Map: Midwest and Western U.S. (justinobeirne.com)
153 points by admp on Nov 18, 2019 | hide | past | favorite | 62 comments



As a long time rural:

They need to be smarter about how small of a road they show at a given zoom level, maybe based on density. If I've gotta zoom in to the point that I can't see the next town over before I can see the road that goes there your map is kinda useless.


I really love these posts because it shows methods of critical thinking about things that aren't strictly lines of code -- and ways of mentally deconstructing things. The 1st and 2nd order steps in reverse engineering processes, comparisons at different levels, and so on must take dozens of hours per post and a huge archive of screenshots from select areas preselected months ago for later comparison.

I also found the general "greening" of the map at larger scales interesting -- something that seems to be a drop in detail -- while the green spaces at smaller scales seem to be more specifically defined (and the new maps seem to have more of them).


Ask HN:

Do Apple Maps, Google Maps, Yelp, etc scrap local business listings? Local search still remains very frustrating. I'm baffled how stores, dr offices, public parks which I know exist don't appear in results. If I was doing the work, I imagine I'd start with every known address and every known business, try to pair them up, and then triage the remainder. Classic data quality efforts. No? I recall an OpenMaps presentation on measuring data quality maturity. Much like bug curves, where rate of changes is used to estimate completeness (coverage). Commercial map vendors do the same, right?

More selfishly, I'm always looking for places to run with my dog. I use a combo of Apple Maps, Google Maps, and AllTrails. I have no idea how local governments account for open spaces, but I imagine they're all in a GIS somewhere, and that Apple & Google have data feeds.

Also local data search (on mobile), if there's a way to geofence results, as in only show shit within my timezone, or even within driving distance, I haven't stumbled across that feature. Driving thru Phoenix, I don't want listings from Indiana, ffs.


Map licensing is very tricky. Typically "open" government mapping data is incompatible with both openstreetmap and free Google/apple maps.

Also, governments tend to be bad at mapping things they own. The guy who mows the grass knows where it is, so they don't spend money to make a map of it.


US states and counties often have extensive data that is available under what is basically a disclaimer. They don't really analyze or assert copyright. Google sucks that data up for sure. They buy data from some cities and counties too.

It's hard to push through imports in OpenStreetMap when all you have is a disclaimer, but the data licenses are probably mostly compatible (public records laws requiring the release and things like that).


What happened with that chunk of land in Lake Michigan that appeared out of nowhere on the new map?


Insider here. Apple pulls in map data from all kinds of sources. Some of these sources exist in alternate timelines. Apple maps will be a lot more useful when the apple timegate finally launches in 2038.


This comment is too good for a mere upvote.


Don't worry, Titor will upvote on his next pass through.


Maybe wrong interpretation of lake ice.

The Upper Peninsula of Michigan is often the victim of mapping oversights.


Maybe, but the map error is too large. There would have to be 20+ miles of ice from the shoreline. And why isn't there ice anywhere else?


Nothing unusual about high ice cover there.

https://www.glerl.noaa.gov/data/ice/historicalAnim/


that's awesome, thanks for sharing.

It's amazing how much variability there is between years


How can they possibly miss that? I want to like their maps, since they browse so quickly on my phone, but it's shit like this...the content is just bad.


I was wondering the same thing, they even got rid of an entire lake.


Yeah that... definitely does not exist in real life.


How is that the Midwest without including MN, WI, IA, and MO?


The headline got this Nebraskan’s hopes up. The term “Midwest” should be retired.


I'm from Michigan and I prefer "Great Lakes region" when talking about the area described in this post.


But the problem there is that the "Great Lakes region" typically includes Wisconsin, Minnesota, and Michigan's UP. And none of those were included in the update.

So it's not really any better than describing the area as "parts of the Midwest".


I was thinking the same thing. Those states are more populated than southern and Rocky Mountain states.


I bet maps rocky mountain states get used more, though, as folk don’t go out of their way to see corn fields the way they do mountains.


No but the people living in Midwest cities do use maps and they outnumber the Rocky Mountain tourists.


I'm not sure about the usefulness of their coloring scheme as there seems to be far too much green that isn't a park or public space. For example, I looked at my neighborhood and they have all the area around the on and off ramps of the local bypass colored green. Technically, yes, that area is all grass but I don't think coloring it green is useful.


I'm curious how these maps are made. Are they done by hand with careful planning beforehand to get the "before" shots? Or is there some API that you can call?


Looking at the changes around Midewin National Tall Grass Prairie... I don’t think this is an improvement. AFACT the boundaries between federal, forest preserve district( which has different usage restrictions), and private land is now totally indecipherable.


Well, to be fair, that data was probably was only unintentionally there before (and maybe not actually labeled?).


Is there a normal browser website for Apple maps, or plans for one? I often use Google Maps website first, then "Send directions" to my phone. If there's no Apple maps website, then I can't do this.


There is a Maps app on MacOS. It can send locations to your phone (or your friend's phone).


You can access it through DuckDuckGo.




This redirects to google.com/maps


Interesting. Platform?


It does redirect to gmap on Chrome 78 / Windows 10 (1903) for me.

Status Code: 302 Moved Temporarily

Location: https://maps.google.com/?t=m&ll=33.805464,-84.145363&q=Stone...


Either Chromium or Firefox, on Linux.


What is going on with Michigan's Upper Peninsula? Did anyone at Apple look up there and notice that they paved Lake Michigan?


Maybe there was ice cover during the relevant satellite passes? But I'm not sure why they wouldn't also have issues with Superior too then. (And I'd expect summer passes to be required for things like grass cover, so it's not like they don't look at that)


I've long had a mild streak of map-geekery in me. But I never really appreciated cartography, especially computer cartography, until a couple of years ago when I was put in charge of a mapping project.

Bottom line: maps are hard!

Especially when it comes to colors. And even more so when you apply those colors across a large and diverse geographic area.

Even harder than rendering digital maps, though, is searching digital maps.

I get frustrated with Google and Apple and all the other maps like everyone else. But I never appreciated how hard these things are until I had to wrap my own brain around it.


I love these articles, really insightful into what goes into making a map.


This blog is a treasure of well-written, highly curated articles. The author is very details focused. I find myself reading all of it, while I haven’t been a map enthusiast at all.


Very well written. It was odd how the article mentioned an improved rendering of Chicago's second largest building, and hyperlinked to the wiki page about the tallest buildings in chicago (spoiler: it's the trump tower). Why not just refer to it by its common and well known name?


They did in the title of the before/after graphic. My guess would be that they wanted to add emphasis on the fact that it's the second tallest building, making it seem like more of an omission. I.e how could it be missing the second tallest building in Chicago!?


I'll buy that, thanks.


Would anyone who doesn’t live in Chicago know there was another Trump Tower there? I for one thought all but the NY one had their names changed because the licensees no longer wanted to be associated with such a toxic brand.


It's the 7th tallest building in the US. It's not an obscure building in the middle of nowhere.


I find the animated images that constantly flip back and forth between "old" and "new" images distracting and make it very difficult to understand the changes.

If the author added a button to allow the user to toggle the two images I think users would be appreciative.


Interesting that they speculated that it was an algorithm that was trained by comparing a manually drawn map with imagery. US cities are all fairly uniform as they are pretty much all based on a grid, but they vary in colour significantly from Oregon to Arizona. I wonder what level of granularity they would need to go to in Europe? Is London different enough from Paris that an algorithm trained on Paris would not work London? In Britain you would need to recognise different cultural structures that you wouldn't necessarily get in Paris, one example is many British parks have a raised up circular bandstand for a local brass band. So they probably need to recruit quite specific people for each locality.


Pray you never visit Boston.


Questions,

Do Apple now owns the Mapping Data, along with Live street views? Where previously Apple were just buying Mapping Data from multiple sources and glue them together.


I believe this is the case, yes. Apple still has a licensing deal with Yelp, but I'm not sure if that includes business locations.

Apple has put significant effort into gathering the data for this new map (4+ years).

I'm sure they still include data from other sources as well, that they may not fully own (like satellite imagery).


Also OpenStreetMap and actively improving OSM data around the world https://github.com/osmlab/appledata/issues/154


They still credit Tomtom quite prominently in the iOS app.


Were stop signs and stoplights supposed to be part of this latest update? If so, how do you get the maps app to show them?


I found it surprising that Hawaii was chosen for the first expansion. Maybe someone looking for an excuse to take a trip?


Until recently they were using old TomTom data for Hawaii. Really old crappy data c. 2006-2011. Hawaii was done early because they wanted their own data.


So, what's next? Central for land area or Southeast for population?


Still fails the Malta to Minturn benchmark over Tennesse Pass along US HWY 24 in Colorado. Needs to have Mitchell, Pando, Red Cliff, Gilman visible at the same time at the right places also. Does not. Busted. Go home.



I don't live there, why should i? This one is now burnt as a benchmark, but i have others, more relevant to my location. I've only chosen this example because it was the one where i noticed how badly common online maps differ from what i'm used to from old paper maps, while following up some old railroad history in the Rockies. With the exception of Bing Maps they create some sort of virtual reality, where some places either don't even show up as separately named area, or the labels are wrong, up to some miles away. Not always, but often. This makes them unusable, or at least annoying to use. By that i mean a general overview of where is what related to its neighbours. As it is now, i get streets, POIs , Businesses, whatever else, for themselves, often missing the right name of the area, be it suburbs or other (sometimes historic) places like i'd explore some blank spots. Which defies the meaning of map, i think. Of course, if all you need is to get from point A to B via maybe C with total disregard for flyover country, industrial wasteland, monotonous suburbia while listening to some audiobook, then this is unimportant.

Edit: Warping spatial perception is what i mean!


Why should you correct the map if you don't live there? Because you seem to use the map and think it should be more accurate. I don't see why where you live is relevant.


Sorry, for being unclear. I wrote that i used this as a benchmark/test to derive the general quality of the map provider. Which means if i see that single example as wrong/insufficient, chances are the rest is too, because whatever process(es) produces the maps is somehow skewed. What is the benefit of correcting that single thing when the rest of the data remains skewed?




Join us for AI Startup School this June 16-17 in San Francisco!

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

Search: