more fixes to multitask dataset. Seems stable.

remove breakpoint

semantic mapper script and multitask reader fixes

deleted semantics statistics (useless) and updated the reader with various fixes

updated notebook

delete old semantic_mask2former_swin_mapillary_converted files and use a semantic mapper directly

Seems to work. Implemented SemanticMapper in DronescapesRepresentation. Not fully generic, but good enough.

almost done

refactor multitask gathering to support dependencies

readme updates

mask2former coco/mapillary

rm this too

remove m2f in test set

two mask2formes and one marigoldr

added newer representations for VRE

updates to task statistics

updates to depth representation

slight fix

added task statistics, update collage

refactored all code and will soon include global stats as well

updates to depth_ufo top be (h,w,1) not (h,w,1,1)

Merge branch 'main' of hf.co:datasets/Meehai/dronescapes

notebook update

small changes to dronescapes_reader. Added initial support for norm/std

files_per_repr_overwrites and and hsv support. Other backports too (opticalflow/ndims)

loggez

loggez

--n_workers to eval script bcoz y not

Merge branch 'main' of https://huggingface.co/datasets/Meehai/dronescapes

update to script

check dirs

very small changes

fixed paper number for safeuav

added nparams to f1 score

updated benchmarks, added safeuav baseline

fix eval script

added old evaluation script

squeeze

no need to be argmaex

detach when u want

added collage and updated readme

npz files for edges_gb and depth_ufo

lel

little fixes and added depth_ufo plus edges_gb

sotseg full

val set softseg

softseg support

some readme

update to sample

little update to the reader