Changelog#

Changelog#

2.2#

2.2.5#

Functionality

  • Support converting single-animal projects

  • Support detecting and converting multiple pickled label files

2.2.4#

Others

  • Migrated documentation to GitHub Pages

  • Use pip-tools for compiling dev dependencies

2.2.3#

Bug Fix

  • Bounding box calculation ignores the non-visible keypoints

Others

  • Reduced package size

2.2.2#

Bug Fix

  • Override class names should work with str and list[int]

2.2.1#

Bug Fix

  • Default class names are correctly defined

2.2.0#

Functionality

  • convert() no longer takes test_paths as they are not used in YOLO data.yml.

  • convert() no longer takes list of dataset paths as there is not much use case.

  • convert() now extract class names from dataset config as the default.

  • Added a few argument checkers for convert().

Performance

  • Minor improvement by reimplementing some argument checking.

Others

  • Removed unused internal functions

  • Added docstring for get_flip_idx()

  • Reduced package size

2.1#

New Features

  • convert() can convert single or multiple datasets

  • convert() can help generate data.yml for YOLO model

  • Added get_flip_idx() function

Bug Fix

  • Fix deeplabcut2yolo module docstring

2.0#

Performance

  • Vectorization of convert()

New Features

  • Automatic detection of dataset structure. Now, convert() only needs one argument.

  • Verbosity and progress bar implementation for convert(). Progress bar is optional, need tqdm.

Functionality

  • No more dependency on pandas. Added numpy and pyyaml dependency.

1.1#

1.1.1#

Functionality

  • Return DataFrame when using convert()

1.1.0#

Functionality

  • Reimplement the convert function for more class and keypoint flexibility

  • Reimplement type hinting for extended compatibility with older Python versions

1.0#

New Features

  • convert function