You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Currently input files are specified in the parameter classes for the trackers. This are a number of issues related to this:
There is no standardisation across trackers.
They must be defined in the parameter classes. This will make it difficult to define default parameter sets (Create default parameter sets #175).
Trackers often check that the file exists when they are instantiated.
This complicates testing as they must be created or mocked.
It has prevented instantiating the tracker objects in load_tracker_metadata.
It puts somewhat useless data in the output metadata (it should be superseded by actually propagating the metadata from the input files)
It makes concatenating the output files in batching more complicated.
The input files are also not really relevant to the configuration of the tracking algorithms, so I propose instead passing them as an argument to run_tracker (and other appropriate functions such as detect). By standardising this it will allow the input files to more easily be passed in the batching.
This is a relatively significant change to the core UI so I would suggest the next release version to be 1.0 to align with semantic versioning.
Currently input files are specified in the parameter classes for the trackers. This are a number of issues related to this:
load_tracker_metadata.The input files are also not really relevant to the configuration of the tracking algorithms, so I propose instead passing them as an argument to
run_tracker(and other appropriate functions such asdetect). By standardising this it will allow the input files to more easily be passed in the batching.This is a relatively significant change to the core UI so I would suggest the next release version to be 1.0 to align with semantic versioning.