[Tagging] automated edits seem to remove crossing=zebra drastically

[Matthew Woehlke]

On 17/09/2020 10.07, Shawn K. Quinn wrote:
> On 9/17/20 08:15, Matthew Woehlke wrote:
>> It's also atrocious because it can *only* be verified by survey. As
>> much as we prefer surveys, the reality is that a lot of mapping
>> happens just from aerials, where crossings (both marked and, in some
>> cases, unmarked) can be seen, but signals cannot.
> 
> I have mapped many traffic signals (and, for that matter, stop and yield
> signs) based on shadows visible on the satellite photos. If you look
> carefully enough (Bing and Mapbox Satellite at least), they are there.
> (Local knowledge helps too in some cases.)

*Traffic* lights I can buy. I am more suspicious of the claim that you 
can tell whether they have pedestrian crossing signals or not, or that 
you can reliably identify other signage based solely on outline. *Maybe* 
if you get lucky and have a very clear shadow at the right angle, but if 
you try to tell me you can identify 
https://www.openstreetmap.org/node/7695704414 (n.b. a yield sign) from a 
shadow in aerial imagery, I am going to be deeply suspicious ;-).

-- 
Matthew