by pininja on 2/17/21, 8:54 PM
This is an incredible visualization! Thanks for detailing how you tiled their huge dataset too. Packing the data into RGB channels as you did is really powerful - I've only seen this done for elevation tiles. Do you have a code link to how you're performing the cursor-brush aggregations?
This seems like it could be an awesome application of an cloud-optimized geotiff (COGs) for serverless tiling. I'm curious if you ran across this tech in your research?
I'm not sure where your project will take you, but I'd encourage you to continue! I got a lot of exposure to the vis.gl community when I worked at Uber, and still contribute - Here are some relevant links you may get ideas from.
COG demo: landsat8.earth
GPU tile processing: https://kylebarron.dev/deck.gl-raster/overview/
Elevation tile decoding (also uses workers): https://loaders.gl/modules/terrain/docs/api-reference/terrai...
by meetups323 on 2/17/21, 10:30 PM
Crazy to see just how dense Europe is -- I've been a few times and always assumed I was just "in the popular areas", that's why there wasn't any open land. But no, CA for instance has many swaths of <1k towers at the zoom level I tried, and to find similar density in Europe required going all the way over to rural Turkey/Belarus/Ukraine region. Pretty much all of the western European countries were entirely 10k+ at that same zoom level.
by teeray on 2/17/21, 7:41 PM
I'm fond of
https://www.cellmapper.net/ for a more local view. It's helpful for answering questions like "why do my calls always drop when I drive into region X?" and "why is coverage shit in this area?", also "what cell carrier should I get if I plan to move to X?"
by mxfh on 2/17/21, 7:58 PM
by paulgb on 2/17/21, 7:12 PM
Using a WebWorker as a tile server for more compressed underlying data is a clever approach!
Are you doing anything special to compute the totals in real-time, or just summing over the entire selection each time the cursor moves?
by rsiqueira on 2/17/21, 9:10 PM
by addled on 2/17/21, 8:40 PM
40 million... I knew there was a lot, but didn't realize that many. Another comment says closer to 5 million physical towers, still a lot.
A lot of these towers have GPS receivers for clock syncing as well, don't they?
Back in college I had a geology prof who was using GPS receivers planted in one spot to measure seismic / tectonic movements from one year to the next.
That was over 10 years ago, and I never looked into it much more, but seeing all those dots reminded me again.
I've wondered what kind of resolution they could model with data from the hundreds of cell towers in the area vs the handful of stations they maintained?
by Aardwolf on 2/18/21, 11:06 AM
by tumblewit on 2/17/21, 7:07 PM
makes me realise the internet is so powerful that the idea of 'voice calling' has now completely changed to 'data exchanging' devices. I mean if you think about it cellular will soon lose its 'cell' meaning. The idea of phone numbers might not go away but everything will likely be IP based which means it doesn't matter how your packets are routed technically.
by BigBalli on 2/26/21, 4:51 PM
Great job with the visualization!
Both design and process are top notch not sure why it didn't gain exposure when you first shared it months ago.
I've been working with cell towers for over a decade thanks to https://FindTowerApp.com
OpenCelliD is indeed a great resource but it is important to note that the rows in their db reflect cells (not towers, towers have multiple cells) and they are based on sampling (ie coordinates are calculated based on samples).
Not 100% accurate but like a said, an amazing resource.
by h0nd on 2/18/21, 3:24 PM
In my area the data is definitely wrong. I think the map rather shows samples from users than the cell towers itself.
Also, 5G is missing.
by pheelicks on 2/17/21, 10:12 PM
Great visualization and approach with compressing the tile data. Do you have a comparison of how much smaller the payload ends up being compared to simply sending PNG files?
I use PNGs to encode elevation data in my 3D mapping library (https://github.com/felixpalmer/procedural-gl-js/) and this does a pretty good job of compressing the data, for example in the ocean the PNG files are also very small as the image is mostly black. Different use case I now as your data is much more sparse, but I wonder how close the PNG compression would be compared to your approach.
by sajforbes on 2/17/21, 11:12 PM
by capableweb on 2/17/21, 8:15 PM
I'm not sure how this was done, or how cell towers really work (or where they exist) but many of them appear to be in lakes, rivers and so on. What's wrong here, my understanding or the map?
by NovemberWhiskey on 2/17/21, 6:50 PM
This is cool; one note - I think there's a very visible typo in the visualization - the legend refers to '(3G) UTMS' - should be 'UMTS' I think?
by vilius on 2/17/21, 8:39 PM
When visiting USA for the first time I was surprised how often I would be in a no reception zone. Drove just away from Miami to Everglades National Park - no reception for miles. Drove from SF to LA via Highway 1 - no reception for miles. Being from Europe I just took cell coverage for granted and always have assumed USA has the same.
by 0xFFC on 2/17/21, 9:58 PM
This might be a dumb question, but why is it way lighter than the US or anywhere in the world? Why do they need to install so many towers? Is there anything specific about Europe's geography that requires European countries to install that much tower?
by clort on 2/17/21, 8:19 PM
This looks pretty cool
However, I have an issue. I'm looking at this area (south coast UK) and there are many dots in the sea which are clearly not cell towers?
by Dylan16807 on 2/18/21, 9:32 AM
So instead of 87GB, how much space do the tiles actually end up taking after compression?
I'm also curious what the average size of non-empty tiles is.
by jemurray on 2/17/21, 7:22 PM
This would make a neat piece of art on the wall. I love thinking of the intersection between technology and neural-pathways, this reminds me of that.
by mdasen on 2/18/21, 12:29 AM
Looking at my area, OpenCellID seems to have a lot more data than CellMapper.net (which people often point to). It definitely has way more data points than there are actual towers.
A lot of them are 310-260s so maybe a lot of people have T-Mobile CellSpots hooked up to their WiFi and that's what is getting detected? If you know more, I'd love to know.
by Krasnol on 2/18/21, 7:22 AM
by 0_____0 on 2/17/21, 9:30 PM
Interesting to see the different "styles" present in different parts of the world. Europe is dense with very wide coverage, Japan has regions of very dense coverage that quickly give way to large voids as you get into the countryside (sharp contrasts between urban and rural), and North America is somewhere in between.
by auraham on 2/18/21, 1:42 PM
by nicolaskruchten on 2/17/21, 11:42 PM
by mrtksn on 2/17/21, 7:36 PM
Okay the world looks extremely well connected. What is Starlink's potential user base? Are there a lot of people in the dark parts of the map?
There's this tendency of maps showing something about humans actually being a population maps simply because the stuff displayed happens where human activity happens.
by mssundaram on 2/17/21, 7:07 PM
The link to OpenCellid is interesting
> Locate devices without GPS
So I guess they offer triangulation between towers to find where a device is?
by 0898 on 2/17/21, 10:37 PM
In the UK, the only place I could find with no cell sites is the Scottish island of Eigg.
by wiredfool on 2/17/21, 7:19 PM
I’m really curious how much space a png rendered set of tiles would be. He mentions that the oceans compress well, but the usual method is to simply not render detailed tiles in ocean regions.
by conradev on 2/18/21, 12:12 AM
by Karawebnetwork on 2/17/21, 8:33 PM
This is beautiful data visualization and would make a great metal poster (e.g. displate). Have you considered selling merch made with this data? I am sure it would sell.
by ape4 on 2/17/21, 7:35 PM
CDMA is only in USA and Japan from my informal mousing around.
by samizdis on 2/17/21, 7:17 PM
Is 5G subsumed under 4G for this visualisation?
(Really pretty, by the way.)
by bullen on 2/18/21, 9:04 AM
I get {"status":"error","message":"INVALID_TOKEN"} when trying to download the data.
by mraza007 on 2/17/21, 6:34 PM
Really interesting visualization and I definitely love the map. I am amazed the way you handled such large amount of data
by k__ on 2/17/21, 11:09 PM
lol
Germany beats all countries in terms of density but South Korea, but we have the worst prices for mobile Internet connections.
by ethagknight on 2/17/21, 7:30 PM
This is great, really interesting visualization.
Surprised at how unlit China is, due to restricted data?
by flemhans on 2/18/21, 8:17 AM
Why are there some at sea (not on an island)? E.g. around Denmark. Windmills?
by zwieback on 2/17/21, 7:20 PM
Awesome map. Rural west of the US is still pretty spotty
by jahbrewski on 2/17/21, 7:49 PM
Damn. Humanity really is just one large brain.
by h1fra on 2/17/21, 7:11 PM
wow amazing viz. More than 200K in Paris area this is insane !
(nb: this would deserve a more granular zoom or shape drawing)
by breck on 2/17/21, 6:56 PM
This is incredible. Thank you!
by _bax on 2/18/21, 11:27 AM
North Korea has no phones..
by Thaxll on 2/17/21, 9:51 PM
Looking at Canada is funny.
by ssss11 on 2/17/21, 9:57 PM
Wow look at south korea!
by just_steve_h on 2/17/21, 8:32 PM
by m00dy on 2/17/21, 10:26 PM
great work!