In order to produce interesting results, the Mapillary dataset labels have been aggregated into larger classes, during a specific data preprocessing.
However we could argue that it is not related with the original scope of the project, and just consider "Mapillary_aggregated" as a distinct dataset from "Mapillary". This will cause a simplification of data repository (by removing "full/aggregated" clause), and paves the way for new aggregated datasets, as additional use cases.
Expected actions:
[ ] remove agregation preprocessing from the project
[ ] simplify the data subrepositories
[ ] extend the simplification to the source code, by removing "full/aggregated" clauses
In order to produce interesting results, the Mapillary dataset labels have been aggregated into larger classes, during a specific data preprocessing. However we could argue that it is not related with the original scope of the project, and just consider "Mapillary_aggregated" as a distinct dataset from "Mapillary". This will cause a simplification of data repository (by removing "full/aggregated" clause), and paves the way for new aggregated datasets, as additional use cases. Expected actions: