I was wondering why there isn’t or if there is, an Open Source (Free, or not) Street View like Google Maps. I like geography as a hobby, like my other hobbies, and found GeoHub which sparked my interest in being a free geoguessr alternative. I just don’t understand why Google is the only Mapping service people go to, is Google the only one that has street view? is this a hard business model to achieve?
Please, I’d love too know :)
Gathering street view like data, processing it, running the service and keeping it up to date would require millions if not billions of dollars/euros in capturing equipment, server farms, wages, electricity, lawfirms.
If you want something thats compareable to the google service, its entirely impossible to do it with a project structure like osm has.
Even if someone had the funding to get started, they would most likely get bought out by google or apple or microsoft very quickly once they got comparable results.
Not necessarily. Strap a $300 camera to your roof, use a properly-written app, and start driving and uploading.
Not likely but also not necessarily cost-prohibitive.
Well, if you have a look at some examples that @[email protected] posted in a comment here, you can see that’s not impossible and that somebody is already doing it.
any sane people would same the same about a map covering the whole world, and yet there is openstreetmap.
yes there are many challenges, but if you start small and grow from there it could work and maybe span a town or two in a couple of years…
any sane people would same the same about a map covering the whole world, and yet there is openstreetmap.
No sorry, that’s a completely different amount of data. Look at how large (or rather, how comparatively tiny) the OSM data set is. Now see how much photo data you’d get for that amount.
you have a point there.
and yet we have the internet archive… so it seems to be possible.
Even the internet archive is nothing in comparison to the image data used for street view.
Its of course totally “technically” possible, but it would require some veeery generous donations from some pretty rich people.
Even if you get people to use their phones to just record everything around them, geo-tag it and upload it. All that data would still have to be stitched together by some big ass GPU cluster that does things that currently only big tech companies can do properly at scale.
Even the internet archive is nothing in comparison to the image data used for street view.
honest curiosity, don’t want to flame war: do you have numbers for that?
stitching is no longer a requirement because of 360° cameras, is it? it could also be made on the client side if really needed. if people can use josm to contribute to osm, they can use some other software for stitching?!
have you seen that the internet archive has also quite high res books scans and videos?
if your aiming for covering every small street of the whole world tomorrow, you are right: it won’t work. but nothing would stop to start with a single city or a region?
lets agree to disagree :-)
Is the current generation of street view still just snapshots from different positions with a 360° camera? I thought it was proper 3D scans with images mapped onto it by now. I admit i havent actually used it in years.
But yeah if its just isolated 2D images then its probably not as much as i thought. The processing would still be tough i think but i dont know enough to even guess that properly.
I think doing small scale demonstrations would be cool. The community would probably be able to learn and improve from it a lot and eventually it could be scaled up.
The one from Apple definitely feels 3D while moving cam around, but I guess this is only afterFXs
If it was encoded as super low framerate 360deg video files and you just seek to the place in the stream that matches the coordinates, it could use kits stage and bandwidth than other solutions, but it’s still going to be a lot of bytes.
Actually, there is. There is mapillary (bought by Meta at some point) and KartaView (I don’t think it is that popular). Code is not free, but licence for images is permissive. My personal favorite is Panoramax - completely open source, open licence and even federated!, but it is still in early stages, but keep an eye on it, looks promising. As always, problem with these efforts is not source code, but rather data.
At quick glance Panoramax looks interesting (I have looked some at the others in the past) but lots of the website is in French even selecting English. Maybe that’s something I can assist with.
Yeah, if you use OsmAnd it shows you mapillary images in the app. You can click around the same as with Google, but the images are all open source.
Thus problem has already been solved :)
Damn, panoramax is impressive. They already have the street where I live!
KartaView went proprietary, and they were dishonest all the time. Look at their github issues with more comments. IIRC there were also bought out, by something that’s only popular in some country.
Well, if you want to do it properly you’ll have to mount a 360° camera on a car and drive a car through every street of the whole world. Takes a while. And then you’ll have to obey the countries privacy laws and at least blur all faces and licence plates.
There are other mapping services that do it but Google were the first who got really popular and did it on a massive scale.
And once you are finished you can probably start over to stay up to date
Google doesn’t do that either, so we wouldn’t have to either. I could name a few places that use photos from 10+ years ago, which look drastically different from the present
They definitely update the photos, just faster in some areas than others, visibly.
Depends on the place I guess. I’m pretty sure some are updated
Yeah they are
Google absolutely update most roads every year or so. Busy roads multiple times a year. Even my cul de sac in a minor town has photos every 5 years
I’m sure doing that wouldn’t lead to insanity
https://wiki.openstreetmap.org/wiki/Street-level_imagery_services#With_photosphere_support
I was also looking into this recently, and I think one of the limiting factors would be the storage and processing power you’d need to blur faces and license plates and then let everyone have access to millions of photos on demand.
I have a pretty weak computer doing real time face detection, so I can’t imagine that being the problem. You could literally do it in real time while biking around.
