Feature Request: Outlier Filtering #4
Description
I would like to suggest a feature: Cyclists should be filtered to remove artifacts and outliers.
Why? In some cities there are frequent bike rides with many cyclists, however it is often difficult to see the real mass due to many artifacts. People that have already left the mass, people still riding towards the mass, people somewhere else, watching the mass without using the Ghost mode, etc.
Therefore I suggest to implement filtering to remove the outliers.
Assumption: Outliers are cyclists not close to many other cyclists.
The classification should be based on two numbers:
- How many cyclists are active in a short range around a cyclist? Like 100-200 Meters.
- How many cyclist are active in a long range around a cyclist? Like 4000-8000 Meters.
Depending on these two values, cyclists can easily be classifies as active or inactive participant.
Scenarios to consider should be:
- Tiny mass with small group of cyclists (like 2-7) within long range: all cyclists should be considered active.
- Bigger mass with more people (8 or more) within long range: only cyclists with a certain number of cyclists within their short range should be considered active. e. g. an active cyclist needs two more cyclists within its short range.
Optional: The filtering could be enable/disabled within the client apps, however it should be active by default/when rolled out initially. Probably it makes sense to do the filtering within the backend and only change visualization on client level.
Screenshots from the actual android app in Hamburg in May 2025 with many artifacts. The Critical Mass is barely visible:
Screenshot from a Proof of concept implementation with a classification result:
