>One of the key functionalities required for such a project to be useable in >countries with developed privacy regulation is the >ability to automatically >pixelate relevant parts of the images with a high degree of reliability. It >took Mapillary literally years >to get that nailed down and bring it to the >level of functionality it is at now.
>Which is one of the reasons why, way back when Mapillary started, I was >sceptical about the sustainability because the >part of the product the >detection is required don't have a real associated revenue stream (except if >you a google, or ... and >can use it in one way or the other to sell ads). >In any case doing that from scratch would be a real pain. I believe the OSC >stack is now actually all OSS which would be a >far better starting point -if- >sustainable funding could be built around the whole thing. I've looked at the OSC github repo from time to time. Bit hard to see exactly what's happening but I understand it's now owned by someone other than Telenav, It _looks_ like the latest commit is client-side only. Strugging to find any server-side code there. However, if you go back to commit 1, there appears to be a fully open-source application with a PHP back end, though I haven't analysed the code in any detail - I can just see it's got database interaction in there. OSC I think only allows you to navigate along an uploaded set of photos, or at least it did last time I looked. Maybe the way forward (particularly given both projects are PHP-based) would be to merge some of the stuff I've been working on in OpenTrailView with the first commit of OSC. The main question that needs to be asked for now I think is: is there sufficient interest in developing a fully open-source StreetView-like application within the OSM community, and elsewhere, to make such a project worthwhile? In terms of the critical privacy issues, there are some interesting projects on GitHub regarding number plate and face blurring, for example https://github.com/understand-ai/anonymizer No idea how good it actually is, but I have a number of panoramas with both faces and number plates so I have material to test it with. Maybe OSC have done some stuff here, haven't looked I have to admit. Nick ________________________________ From: Simon Poole <si...@poole.ch> Sent: 19 June 2020 15:06 To: talk@openstreetmap.org <talk@openstreetmap.org> Subject: Re: [OSM-talk] Facebook acquires crowdsourced mapping company Mapillary Am 19.06.2020 um 13:47 schrieb Nick Whitelegg: (Disclaimer: I am the developer of said project) One of the key functionalities required for such a project to be useable in countries with developed privacy regulation is the ability to automatically pixelate relevant parts of the images with a high degree of reliability. It took Mapillary literally years to get that nailed down and bring it to the level of functionality it is at now. Which is one of the reasons why, way back when Mapillary started, I was sceptical about the sustainability because the part of the product the detection is required don't have a real associated revenue stream (except if you a google, or ... and can use it in one way or the other to sell ads). In any case doing that from scratch would be a real pain. I believe the OSC stack is now actually all OSS which would be a far better starting point -if- sustainable funding could be built around the whole thing. Simon PS: naturally the whole reason for OSC was a business dispute that is now moot because Mapillary is opening up its images for commercial use too. Those of you looking for 100% FOSS software and who are focused on 360 degree photography of off-road routes (walking trails and so on) might want to consider OpenTrailView (https://opentrailview.org). Do bear in mind that it is in the early stages of development, so don't expect Mapillary-style UX just yet, and there is only a small amount of imagery (largely southern England at the moment plus a few around Heidelberg for probably obvious reasons) but it is in active development and I do have a possible collaboration with another project (more on that later). OpenTrailVIew also uses underlying OpenStreetMap data to auto-connect panoramas, using GeoJSON Path Finder (github.com/perliedman/geojson-path-finder), though, due to server capacity constraints, this only works at present in Europe and Turkey (though requests for other countries welcome, though note that if they are for large and/or highly-populated countries countries such as the USA, China or Brazil I would have to restrict it to a region). You can login using your OSM account. Nick ________________________________ From: Florian Lohoff <f...@zz.de><mailto:f...@zz.de> Sent: 19 June 2020 07:58 To: Niels Elgaard Larsen <elga...@agol.dk><mailto:elga...@agol.dk> Cc: talk@openstreetmap.org<mailto:talk@openstreetmap.org> <talk@openstreetmap.org><mailto:talk@openstreetmap.org> Subject: Re: [OSM-talk] Facebook acquires crowdsourced mapping company Mapillary On Fri, Jun 19, 2020 at 01:21:59AM +0200, Niels Elgaard Larsen wrote: > Paul Johnson: > > Great. How's this affect those of us who trust Facebook about as far as we > > can throw it? > > > Use openstreetcam Openstreetcam is pretty much "disfunct" from my perspective. There are tons of bugs people opened because of their tracks not beeing processing. Same for me. Twitter feed dead for a year. It looks pretty much abandoned since end of 2019 - Since early June serious problems processing tracks and uploads. And for the me focus on Car driveable streets makes it useless. Flo -- Florian Lohoff f...@zz.de<mailto:f...@zz.de> UTF-8 Test: The 🐈 ran after a 🐁, but the 🐁 ran away _______________________________________________ talk mailing list talk@openstreetmap.org<mailto:talk@openstreetmap.org> https://lists.openstreetmap.org/listinfo/talk
_______________________________________________ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
