Proposed import of buildings in Montreal

Greetings,

I am proposing to import the City of Montreal’s contiguous building outline dataset, parsed into individual building outlines and with out-of-date buildings removed using the City’s Property Assessment Roll dataset. Both datasets are sourced from the City of Montreal’s open data portal.

Documentation

This is the wiki page for the import:
https://wiki.openstreetmap.org/wiki/Montréal/Imports/Bâtiments
This is the source dataset’s website:
https://donnees.montreal.ca/dataset/batiment-2d
The above source, containing the contiguous building outlines, has been parsed into individual buildings and has had stale data removed with the help of the following data:
https://donnees.montreal.ca/dataset/unites-evaluation-fonciere
This is a file I have prepared which shows the data after it was translated to OSM schema (conflated with OSM data as of January 21, 2023):

License

I have checked that this data is compatible with the ODbL.
This data is distributed under CC-BY-4.0. Please see the following links for more information:
https://wiki.openstreetmap.org/wiki/Montréal/Imports/Ville_de_Montréal
https://donnees.montreal.ca/pages/licence-d-utilisation

Abstract

The dataset contains contiguous building outlines from the City of Montreal, parsed into individual building outlines and with out-of-date buildings removed. Buildings that intersect with existing OSM buildings were also removed. This import would total just under 137 000 individual buildings. All buildings were simply tagged as building=yes and also include a tag for the source linked above. Conflation, i.e. removing buildings from the import which would intersect with existing OSM buildings, will be repeated once more on the day of the import in QGIS. Using JOSM, the data will then be converted to an OSM XML and uploaded in chunks of 9000.

Next Steps

Please let me know of any issues. If there aren’t any, once the delay stipulated by the import guidelines passes, I plan to proceed with the import.

Thanks!

1 Like

Having recently painstakingly hand-drawn a few thousand buildings in the area, I would be very happy to see the rest of this automated.

Many of the areas with buildings already drawn in are very outdated, based on imagery from 10+ years ago. I have been manually realigning them all based on current Bing (and when Bing is outdated, Esri) imagery, to the correct offset of (0.6,-1.54). This too is very painstaking. Is there any way we could look at updating certain boroughs after the import of empty areas has been completed? (See here for the part of Ville-Marie I have already realigned.)

I haven’t looked into automated updates to existing buildings as much, but I imagine it would be somewhat more complex than this import.

With that being said, I’ll look into it if this goes well. If I do find a way to contribute to that end, I’ll definitely post about it.

1 Like

“une_abeille” is an adequate pseudonym looking at the work done!

1 Like

Can you provide a screenshot of a sample of the level of detail?

One reason previous attempts to import buildings or trace them with AI have been rejected is that neighboring buildings or whole rows of buildings would have been imported as one. For example, see here a basemap used by Strava which uses OSM buildings when possible and AI generated buildings from Microsoft (I think) otherwise. You can clearly see the difference between the buildings I traced west of Papineau and the buildings traced by AI east of Papineau.
image

In my opinion, I’d rather have no buildings than badly traced buildings, because it is much easier to trace them from scratch than to realign and split them. But certainly if the quality is good enough it would be amazing to do this import.

1 Like

Sure of course. Here’s an example of what the building data would look like (red outlines):

Even if buildings share a wall, they will still be separate ways.

Something related to this that I do want to mention is that some areas will still not have complete building coverage. For example:

This is because the resulting buildings fall outside the tolerances I noted on the wiki page. Most of the time, the buildings did generate as expected, but the underlying lots (used to discern individual buildings) are not aligned properly in the dataset. In such cases, I don’t feel comfortable uploading these buildings without first reviewing them manually.

As an update to the original post, the buildings will only contain the building=yes tag.

Initially, I made the mistake of including a redundant source= tag on each building. According to the wiki article for key:source, it is only necessary to include this tag on the changeset.

The example file has been updated to reflect this change.

2 Likes

I’m looking at those missing buildings and thinking it would be a shame to miss this chance to import them. How big a task would it be to review those buildings manually? Can it be done in this changeset?

I’ll try to reincorporate as much geometry as I can find that was erroneously removed for being “glitched.” This will take a while though, so it will probably continue even after import – but it will be added eventually.

Update: I spent some time today reviewing the buildings that were rejected, so there should be significantly fewer empty spaces in the import dataset. The example OSM XML has been updated to the latest dataset.

2 Likes

This test set of buildings looks perfect!

Edit: The whole import is looking great. I’m not seeing too much that’s missing, and I’m sure that with a list of areas that were rejected we can easily manually finish this off. Thanks so much for putting in the work on this import!

As une_abeille noted, everything on my end is now uploaded.

The upload process was interrupted when about a quarter of the data was left to upload due to a rate limit error, but everything was able to be resumed and completed without a hitch after waiting about 50 minutes.

There is still a small minority of buildings left to complete manually.

I’ll continue to look over what was just uploaded.

A big thank you to everyone who helped me with this import. :slight_smile:

2 Likes

One thing I noticed is that buildings in these areas were not imported, likely because there were existing buildings there up until yesterday. I removed them because they were very poorly drawn; basically one large outline for an entire block of plexes. They will need to be manually drawn, or perhaps a smaller import can be done.

I suggest you use josm “replace geometry” (or similar) instead of deleting existing features.

Replace geometry Ctrl+Shift+G : this function allows you to redraw a feature that has been drawn poorly or its real-life shape has been updated while keeping the original feature’s tags and history.

They had no tags other than being labeled as buildings. I also do not have JOSM.

Quelle application utilises-tu pour faire toutes ces éditions?

Nice work @Spearfish7424 This import could have been way way worse! We just need to fix some small issues, which is great! Thanks! (I see you already fixed some weird connections between buildings).

If you guys want to get enhanced landrole data and the official “cadastre”, feel free to ask :smile:
The landrole data includes the number of flats and the address. Most building also have associated total living area and number of levels.
See an example in Ahuntsic:

Since we apparently have permission to import any of the city’s open data now, it would certainly be nice if we started importing other data sets as well. I was considering trying to learn how to do imports with one of the smaller data sets like fire hydrants.

But we should probably polish off this one first. I was going to start finishing some areas tonight where there are empty spaces (from the buildings I mentioned deleting above.) @Spearfish7424, if you need help with any of the cleanup, please feel free to point me somewhere!

Addresses would be very high value to import, and both those and building levels would save a lot of StreetComplete quests.

Thanks @ChaireMobiliteKaligrafy ! The weird separations between buildings should now be few and far between. These were caused by features such as skylights or chimneys interrupting the lines which separate attached buildings. This is something I should have noticed before the import, but it should be fine now.

Thanks for the offer to help @une_abeille. The glitches in building separations should now be fixed. Do let me know if ever you notice anything weird though.

I added the buildings for changeset 147240584 , but the rest are less ideal to import.