Talk:Canada - The Open Database of Buildings
The purpose of this page is to provide a detailed inventory of ODB data, to provide the current status of the import, to describe the quality of the data (in order to assess what it may bring to OSM) and, finally, to allow informed discussion on the steps to follow to complete the import. Importation had already begun when many contributors asked for a broader consensus on how to proceed with the import (pre-processing, integration with existing data, etc.). The import was therefore stopped in order to reach a wider consensus before proceeding.
Overview of Data Quality
In December 2019, a complete inventory of both the ODB data and the status of the corresponding data localities in OSM was carried out. In summary, the analysis shows that the quality of the data differs from one municipality to another. Detailled results are provided in the main page.
Data content: Some offer a complete set of detailed buildings (e.g. Edmonton, Victoria, Vancouver). For others, the level of detail changes depending on whether the building is in a rural or urban environment (e.g. York Region, Cape Breton). Some provide only a partial content (e.g. the remainder of Nova Scotia, Sherbrooke). Finally, and in some cases only, other structures (e.g. wharfs, swimming pools) are included as buildings (e.g. Moncton, Halifax, Rimouski). In most cases, sheds are included in the datasets.
Representation accuracy: most datasets represent buildings’ footprint with a level of detail comparable/superior to what is already found in OSM for each municipality. However, cases of inadequate shapes (e.g. bounding boxes, overlapped polygons) were detected in most of the datasets. Similarly, provided polygons require to be orthogonalized (squared) in many cases, but not all. Towers built on larger buildings seem occasionally mapped as a standalone building instead of being a component of the larger one (e.g. Vancouver, Toronto). Terrace buildings (i.e. a series of buildings that share walls) are represented as a single large building rather than as individual ones.
Positional accuracy: It is reasonable to assume that provided buildings’ location is generally better than what can be achieved with Bing’s imagery (particularly over hilly areas) or with a GPS device (signal distortion around the walls). The one exception is Ontario with Bing’s imagery falling within a meter of building footprints most of the time.
--Jfd553 (talk) 11:30, 9 January 2020 (UTC)
Clarifying questions about measurements
I have a few questions about how these observations were generated.
- What exactly is dx,dy? Is this an offset? In metres? From what to what? Is this an average? If so what kind? How many observations were used to produce the average?
- How was the percent of buildings mapped estimated? There is no perfectly complete dataset against which to check OSM. The ODB data itself is definitely not a valid benchmark. Is this total area mapped, total building count, etc?
- How was the source of mapped buildings estimated?
Nate Wessel (talk) 15:07, 26 December 2019 (UTC)
Comparison method:
1- All ODB buildings were converted from ESRI shape files to OSM format. Separate OSM files were created based on their data provider (Data_prov). A small number of buildings that were not OGC-compliant were removed from the datasets (<0.001). Each resulting file was read into JOSM for comparison. An examination route was mapped over each file using usually 5–10 turning point locations in order to cover all the data.
2- ODB & Bing Comparison: Bing’s imagery was aligned with ODB data for each of these locations to assess the DX and DY values (meters). Bing’s images and ODB data were compared all over the examination path to assess the shape of ODB buildings (i.e. the shapes are just bounding boxes or they provide a very detailed description of the buildings). The completeness of ODB was assessed at the same time. Completeness has been roughly coded as follows: 90% means that almost all of the buildings on Bing’s images are available in ODB; 10% means that only a few buildings visible on Bing’s images are in ODB; and so on as the proportions change. Unexpected discoveries (for example, swimming pools or structures are provided as buildings, overlapping buildings are found in ODB) were also noted.
3- ODB & OSM Comparison: Also in JOSM, ODB data was displayed on OpenStreetMap Carto representation (standard). The same examination route (see point 2 above) was used to compare what ODB offered and what already existed in OSM. The completeness of the OSM data, compared to ODB, was assessed using the same rough coding used above. When OSM and ODB were very similar, the description of the OSM changesets were downloaded and examined to determine whether the data came from ODB or from another source. Unexpected discoveries (e.g., imported ODB buildings overlapping OSM street network) were also noted.
--Jfd553 (talk) 21:54, 27 December 2019 (UTC)
Availability of ODB Data
Obviously, the ODB data is always available from the Federal Open Data web site in shape files. However, since we move toward using the task manager, I wonder if the following information needed to be presented. I removed it from the wiki page but keep it unchanged below just in case...
The data used for this import are released by Stats Canada and will be served through a data service in MBTiles Vector Tile 2.0 format, thus you must provide a z,x,y to obtain data. Building Footprints (Polygons) https://data.osmcanada.ca/{z}/{x}/{y}/statscan-buildings.osm
OSM Data Files
The data will be served by https://data.osmcanada.ca/ as JOSM xml files via a link in the tasking manager. See data source site for more information.
--Jfd553 (talk) 20:25, 11 January 2020 (UTC)
ODB Inventory Tables
To ensure adequate monitoring of import tasks, the ODB inventory tables will be modified as follows:
Added to tables:
- The name of the local import manager (User),
- The link to the task in the task manager (Task Manager),
- The estimated task completion (Completion).
Removed from tables:¸
- The Approximate proportion of actual buildings mapped in OSM, and their origin (ODB & OSM)
- The Import status as (F)ull, (P)artial or (N)one (ODB).