I am proud to announce that my and my two roommate’s fun, totally free, cycling route creation site, https://sherpa-map.com, finally has its own map!
This map ended up costing thousands of dollars to produce, many, many, months, redos, research, etc. to combine everything we could to try and provide all of the cyclist-specific information possible at a glance.
This map incorporates road colors that identify road surface types using both OSM data and roads I have classified by pulling satellite imagery and using AI to determine unknown road surfaces en masse.
(black = paved, gray = gravel, tan = unpaved, brown = dirt, pink = unknown)
Additionally, this map uses data from Facebook’s Machine Learning project Daylight: Roads (Daylight Map Distribution)
Which scans the planet for things that look like roads and adds them, you can’t route on those yet, but you’ll be able to see them on the map to help inform your journies.
The core of the road styling is borrowed from Cyclosm cyclosm-cartocss-style/docs/DOCKER.md at master · cyclosm/cyclosm-cartocss-style · GitHub I’ve heavily modified it to include more squiggly fun roads when further zoomed out, adjusted road size, coloration, etc. I’ve kept a huge emphasis on showing anything and everything bike-related over practically anything else, scenic cycleways, mtb trails, bike trails, etc.
Other than the road coloration differences for surface type, the full legend can be found here: CyclOSM - Legend
I did render this map for the entire world, but, it’s only really usable down to zoom level 16 (quite zoomed in!) for:
United States
Japan
Philippines
Taiwan
Canada
Australia
Europe
Alaska
Hawaii
Other zones are on their way.
Additionally, this is technically two map layers! A road layer, and a hillshade layer, I developed the hillshade layer using the highest resolution Lidar and Satellite elevation data available, I want you to be able to pick out every hill on your route.
I used QGIS, Gdal, and Mapnik to produce this, I also aggressively cache this in Cloudflare and on the browser, so the more you use it the faster it will be.
This is version 1.0! I’ve already come up with many features and changes for future versions, I intend to make many, as we purchased a 64-thread 5Ghz Threadripper, 128gigs of DDR5 ECC, and 48 TB of storage to generate this, so we’re definitely not stopping here.
I have a whole new UI coming, with tons of nifty updates, I have more map updates coming as well (like an overlay that shows, visually, as many gravel races as we could manage to find GPX files for) if you want to follow along for any updates, I tend to post them here: Redirecting...
If anyone is curious to see what it looks like computer-wise to render the 2.8 BILLION image files that comprise these two map layers, loooook at this task manager:
We spent months with the computer pegged like this, we nicknamed it “Hurricane” because it was so loud.
Any thoughts and feedback are welcome, with the expertise I’ve gained creating this, I can essentially do whatever I want with future iterations, I’m not locked into any frameworks and systems like Mapbox/Vector tiles, etc.