Background:
Rotterdam3D is a dataset of 3D buildings which are not only in LoD2.2 but also include information about underpasses and other building features. The underpasses in this dataset were mostly modelled manually. In the 3DBAG project, we are trying to detect them automatically. Therefore, we will use the Rotterdam3D dataset as ground truth for the verification of our method.
Below are some images of underpasses and how they look in the dataset and in reality, along with links to the actual locations. We are especially interested in cases like the last one: buildings with not too many floors that have an underpass towards an interior yard.
Task:
You need to download data from the Rotterdam3D viewer, and see how you can automatically identify where underpasses are. We recommend subtracting roof surfaces and ground surfaces and looking into their differences. Keep in mind that there are usually multiple roof surfaces for a single building. You will probably need to merge those and project them to the same plane as the ground surface of the building.
How to download data:
Data Source: https://www.3drotterdam.nl/
Area of interest: Start simple with a small area where you know there are some underpasses. I noticed that there are quite a few underpasses around the locations of the examples above.
Format: cityjson
The data will be sent to your email.
Where to visualise:
Where to view: https://ninja.cityjson.org/#
Great because different surfaces appear with different colours.
Where to process:
Probably in Python
Expected Output
Minimum: A list of building IDs where an underpass is detected.
Ideal: A list of detected underpasses, each accompanied by a corresponding 2D geometry.
Documentation:
You will encounter many ambiguous cases where it is unclear whether a feature should be classified as an underpass or just a balcony or just a small building feature. This information will be very valuable to us so please document those cases with examples.
Deliverables:
- A python script that allows us to identify the underpasses in the Rotterdam3D that takes as input a cityjson file
- Documentation of the cases where this didn't work and where it worked well and the decisions you had to make (with images).
- An example of a sample cityjson file and the output of your script
Responsible: @GinaStavropoulou
Background:
Rotterdam3D is a dataset of 3D buildings which are not only in LoD2.2 but also include information about underpasses and other building features. The underpasses in this dataset were mostly modelled manually. In the 3DBAG project, we are trying to detect them automatically. Therefore, we will use the Rotterdam3D dataset as ground truth for the verification of our method.
Below are some images of underpasses and how they look in the dataset and in reality, along with links to the actual locations. We are especially interested in cases like the last one: buildings with not too many floors that have an underpass towards an interior yard.
Example 1:
Example 2:
Example 3:
Example 4:
Task:
You need to download data from the Rotterdam3D viewer, and see how you can automatically identify where underpasses are. We recommend subtracting roof surfaces and ground surfaces and looking into their differences. Keep in mind that there are usually multiple roof surfaces for a single building. You will probably need to merge those and project them to the same plane as the ground surface of the building.
How to download data:
Data Source: https://www.3drotterdam.nl/
Area of interest: Start simple with a small area where you know there are some underpasses. I noticed that there are quite a few underpasses around the locations of the examples above.
Format: cityjson
The data will be sent to your email.
Where to visualise:
Where to view: https://ninja.cityjson.org/#
Great because different surfaces appear with different colours.
Where to process:
Probably in Python
Expected Output
Minimum: A list of building IDs where an underpass is detected.
Ideal: A list of detected underpasses, each accompanied by a corresponding 2D geometry.
Documentation:
You will encounter many ambiguous cases where it is unclear whether a feature should be classified as an underpass or just a balcony or just a small building feature. This information will be very valuable to us so please document those cases with examples.
Deliverables: