Great to see new imagery sources! Hope to connect with you, maybe your familiarity with this could aid with future project ideas too

I am not sure I understand when you claim Overture hates copy left? This is your view, and does not represent anyone. I love OSM (dearly) and have worked on some pieces of Overture. I don’t have hate, and I don’t know anyone else who does. Please don’t create chimeras out of people. This just shows again the disruption in point #1: emotional reactions to Overture, regardless of the technical or practical aspects.

If Overtue hates copy left, then it would seem fair to say that in inverse, OSM hates being totally free and open (permissive). But is that really true?

If there are 15 data sources, OSM still remains a good one: sources are worth comparing. Permissive ve is far easier to mix with other data, ODBl has all kinds of things that need caution, you can read the OSM Wiki on this.

All the permissive data is excellent for OSM: it becomes available to go into OSM. Daylight is important to Overture. Overture alles others to choose which data they use, no reason to abandon OSM.

Currently the only I disruptions I see are:

1. Emotional reactions by some in OSM that Overture exists
2. Companies have an ability to download and use some basic map data (roads, buildings, POIs, boundaries) with the Open License in a central place where other companies share it without directly inputting it into OSM first (piecemeal or negotiating an import)
3. OSM has an increasingly large amount of global data that is fully open for it to import or reference to fill in blank spaces, if the data can be verified as accurate and precise

I don’t argue for any particularly noble cause here, but it will serve a purpose and has a net positive for those it impacts, and others who see it as negative, I don’t think it really impacts unless they are operating a business that seeks to make revenue doing something Overture provides for free, perhaps.

The main one that interests me is not here on its own maybe, but this is whether or not pets are allowed in a hotel, a restaurant, etc. I think that is just pets=*?

Wow! Nice work @wille!

This is fantastic research and documentation! I know some experiments have happened with Mapillary imagery in the past in this regard.

What also could be very interested is to track pixels in this process, in order to turn detections of roads, buildings, sidewalks, vegetation etc into 2d vector polygons, since Mapillary already has open data about the image segmentations and labels.

I don’t see this being pursued in depth during 2023 via Mapillary, as the agenda is quite full with other great stuff, but I hope to share this around and see where it can generate interest with colleagues!

@SomeoneElse

1. the confidence scores I personally wrote are not worth trying to use, no reference point.

2. The data quality is variable by region for sure, but so is OSM. I am not sure if we could say it’s lower or higher quality than the POIs already on FB (in my area in Switzerland with a quite active community many are 5+ years old and there is not system nor paid team to maintain them)

3. I am not sure OSM has a definition of quality? Is there any written method of evaluating the quality for OSM?

Bottom line, all POIs are questionable until checked by a person, which is exactly the opportunity these present.

@CjMalone do you mind if I update my JSON dict with some of your tags, then merge the rest of mine into your table? We can keep it evolving until it’s correct.

@SomeoneElse - I wouldn’t call that the biggest challenge at this stage. You can’t start to verify if the data is not connected. If some just do not have a good or easy tag match, it’s already thrown out probably. Validating the rest is not difficult per se (not a crazy math equation to solve), but rather, a long, slow, and careful effort. We don’t need to make sure the dataset itself is not utter garbage, only individual line items. That’s quite a small effort for individual users on a local scale of let’s say a 1km radius of your home/work.

If you think quality is not a factor in FB Places data, yikes.

I do not mention the confidence rating because I do not see a good use of it. Without some kind of survey, confidence is unverifiable to OSM users.

In the end I perfectly agree though: this is useful as an aid to survey. That is well put.

Which AI would attempt to add the data? I have not seen AI make changesets myself. AI is excellent for converting 1000 labels to potential equivalent tags, that for now me, and “someone else” (anybody interested) can help manually check for right conversion, before then looking to check for existing matches on OSM to know what to throw out as already existing, then whittle down to opportunities worth validating.

Are you using Rapid editor to add buildings? Looks like the AI detected buildings from Microsoft are pretty nicely matched with the Bing Maps Aerial imagery too, but would be great to see the WMS you listed get added to the aerial sources.

Are you using to add buildings? Looks like the AI detected buildings from Microsoft are pretty nicely matched with the Bing Maps Aerial imagery too, but would be great to see the WMS you listed get added to the aerial sources

Nice work, that looks great!

What failures of the microgrants program risk making it a waste of money?

Congrats, Russ! Great to have you. I still owe you a beer for helping me try and get a credit card to me in Tanzania even if I screwed it up!

Thanks and welcome! StreetComplete really is fantastic.

I would recommend mapping a few terraces but checking the button that asks for review of your edits. You can also then share a link to the area you edited on the OSM Mailing List, or other groups like Telegram or OSMUS Slack. It’s great to have more people join the discussion and ask questions about detailed mapping methods.

Awesome read from a true veteran! Are you using Mapillary these days for your photos, or just taking still photos manually?