vanvalenlab / caliban-toolbox / 444

85%

push

travis-ci-com

DatasetBuilder and DatasetBenchmarker (#115)

* modified train/val/test function

* refactored init

* switched to loading helper function

* loads all experiments and splits data

* build dataset function

* first pass at DataBuilder

* working version of bencharmker

* removed separate id and name fields

* integrated resizing

* pep8

* switch naming in Benchmarker

* pep8

* kwargs to pass in additional arguments

* fixed mocker

* handle images with no labeled cells

* fix bug with resizing

* corner cases for train/test/val

* incorporate train_val_split changes

* added summarize

* simplified output format for Benchmarker

* simplified _reshape_dict

* renamed variables

* pep8

* renamed files

* fixed bug in summarize

* different output shapes for each split

* simplified benchmarking dicts

* renamed file

* added explicit 4D checks where necessary, abstracted to ... where not

* extract error checking

* bugfixes

* Apply suggestions from code review

Co-authored-by: willgraf <7930703+willgraf@users.noreply.github.com>

* add option to resize by constant value

* add testing for reproducible splits

* pep8

* better input validation

* Apply suggestions from code review

Co-authored-by: willgraf <7930703+willgraf@users.noreply.github.com>

* typo

Co-authored-by: willgraf <7930703+willgraf@users.noreply.github.com>

880 of 880 new or added lines in 8 files covered. (100.0%)

2501 of 2981 relevant lines covered (83.9%)

1.67 hits per line