Improving POI coverage

[skorasaurus]

Goal: Improving POI coverage in the city.

Very briefly will mention that this is just one way to do it and a supplement (alternatives include and aren't limited to having in-person surveying of locals (mapping parties), obtaining open data sets from the city of Detroit, and importing it into OSM) .

Since I'm not local and contribute to Detroit's OSM community and coverage with limited time, I had been using mapillary to add places of interest (businesses, churches, etc) on Detroit's East Side that are open to the public on main streets.

Source data: mapillary photos; the ones that I've seen are from Nov and June 2017.

In my experience, I've found it more productive to do one side of the street at a time so you can ensure that you're not missing anything in pictures by having to go back and forth between different photo sets (since most sequences [uploads of photos] contain photos oriented towards only one side of the street) , and ensure that the objects/POIs that you're adding are on the correct side of the street.

Disadvantages with this approach (not limited to just these):
sometimes the pictures are too blurry, taken at the wrong angles so you're unable to identify the name of the business or even determine the place's purpose (is it a social center, a clinic?)
unable to sometimes know minor nuances (whether a place is more of a restaurant than cafe, etc).

Admittedly, my time to commit on this is limited and cannot commit to leading this.

What I've done so far:
Gratiot Avenue
east 7 mile road (chalmers to redmond street)
Mound Road (west side of street)

Other things to keep in mind:

• Determine whether the community wants to map (use nodes or the entire building for address points) for consistency's sake

A couple example changesets:
https://www.openstreetmap.org/changeset/61063604
https://www.openstreetmap.org/changeset/61058355
https://www.openstreetmap.org/changeset/61093425

[mikelmaron]

This is an awesome effort @skorasaurus. I wonder what's the best way to task this up. Sounds like the ideal unit of task is one block. The OSM Tasking Manager could possibly be put to use, if we developed a custom GeoJSON to define the task: split up the street grid from one neighborhood into block segments, and buffer slightly. Thoughts on this approach?

[jonahadkins]

i looked at this today - starting with the streets & parcels created a layer of "blocks" to potentially load in the tasking manager - added the city's neighborhood boundaries to break them up into groups - its still a lot - i've got a directory of 200 geojson files, 1 for each neighborhood

here's the Corktown section

maybe start with 10 surrounding downtown with 1 task for each neighborhood linked by a hashtag for the challenge?

[mikelmaron]

@jonahadkins this looks great. let's do a starting test with a handful as you suggest. maybe a selection around the city to get different kinds of landscapes covered. we can test out the directions and workflow and task size. please add the geojsons somewhere accessible, and I can create a few tasks.

[mikelmaron]

I created one test task to start. Do the tasks and the instructions look good @jonahadkins @skorasaurus

https://tasks.openstreetmap.us/project/50/dashboard

[skorasaurus]

Great writeup, a couple suggestions:

• Suggest to use MapBox aerial imagery; I think it may be the most recent imagery available (I don't know when it was captured and I am not sure that it is the newest one; confirm with a local before inserting that language)

• Suggest to use OpenStreetCam or Mapillary imagery; the Microsoft streetview is from 2014 in the areas that I've mapped; From what I've heard, there's noticeable construction and demolition in parts of Detroit since then. I defer to your and local mappers' judgment whether to recommend MS at all but suggest to consider it.

• Decide the scope of "points of interests" a little more and specify it; as a potential mapper, I'd be unsure whether I should map a construction site, a vacant building, a house, etc. Would it be more advantageous to instruct users to only map buildings that have a POI on them since the rest of the buildings can be taken care of in an import?

[mikelmaron]

Good points @skorasaurus, will update

[mikelmaron]

Dexter shared this CSV of 34k Detroit POIs, a composite put together from a bunch of sources, needs cleaning up but it's got mobile eyes data (from fire inspections) business licenses etc.

You can see a quick and dirty map of this data

Ideas for this data

• Match POIs against OSM features, identify missing POIs
• Create a background layer on satellite imagery, integrate into task, additional set of info
• Task to systematically check and selectively add POIs to OSM

[iandees]

It's a bit work to trim down to just Detroit, but there is lots of POI information in the http://www.alltheplaces.xyz/ download.