Skip to content

refactor: Change annotations organization

Mathias Réus requested to merge unified_split into master

We merge the annotations CSV - annotations_test and annotations_train - into a single annotations.csv file. Also we added a column : split_pedes. Column type is renamed split_sysu and its values are train, query, gallery. This commit reply to issue #17 (closed).

The refactor passes all test, stats notebooks has been rerun. We also clean some notebooks, so we do not maintain development notebook #11 (closed).

We made a huge change in modules organization

  • sysu_annotations and pedes_annotations get annotations.
  • build module is used to create the dataframe of the annotations
  • io module is used to read/write annotations dataframe
  • coco module is used to isolate boilerplate code for JSON COCO seralization operation
  • constants module groups some constant
  • cli module integrates former build annotations scripts. User can perform merge of annotations using the package directly.

README have been updated and build process also. We added pytest as testing tool.

Merge request reports