This PR aims at simplifying the definition of datasets, by removing the aggregation mode (aggregatedvsfull). This parameter was introduced in order to design a simpler Mapillary dataset, with less classes. One now considers that the Mapillary dataset corresponds to the "aggregated" version, and encourages to modify the dataset name to lighten the preprocessed/output folder structures.
Hence, one may use the following dataset:
mapillary for the aggregated Mapillary dataset
mapilfull for the raw Mapillary dataset
mapilbuilding if focusing only on the building class (respectively to any other class of the dataset).
This PR aims at simplifying the definition of datasets, by removing the aggregation mode (
aggregatedvsfull). This parameter was introduced in order to design a simpler Mapillary dataset, with less classes. One now considers that theMapillarydataset corresponds to the "aggregated" version, and encourages to modify the dataset name to lighten the preprocessed/output folder structures.Hence, one may use the following dataset:
mapillaryfor the aggregated Mapillary datasetmapilfullfor the raw Mapillary datasetmapilbuildingif focusing only on the building class (respectively to any other class of the dataset).