Samantha Prestrelski

Refactor to use pathlib instead of os.path and path concatenation

1

In IsoAutio, when running generate_automated_labels, the path concatenation does not allow easy iteration through folders. It would be good to use pathlib any time we deal with paths to make...

Complement to ROC curves - good for situations with heavy class imbalance (AKA our multi-species classifier training data)

Add function to turn annotations into one hot encoding

1

Turn weakly labeled files into a DataFrame using one hot encoding. | | Class 1 | Class 2 | Class 3 | Class 4 | Class 5 | Class 6...

PyLint

4

Add linting, similar to the setup here: https://github.com/UCSD-E4E/acoustic-multiclass-training/blob/main/.github/workflows/pylint.yml This should also include some restructuring to remove old/irrelevant functions

Instead of showing individual file processing, refactor to use tqdm(https://github.com/tqdm/tqdm) to show a progress bar

Instead of general exceptions, specific exceptions should be referenced Mentioned in #159 https://github.com/UCSD-E4E/PyHa/pull/159#discussion_r1396672210

Some functions currently have assertion statements, but given #159 it would be good to add unit testing for individual functions, especially for empty DFs, path errors, and file reading errors.

This implements BirdNET-Analyzer instead, which is currently being updated: https://github.com/joeweiss/birdnetlib/ It also will decrease the size of PyHa since we wouldn't have to store the model anymore

From Sean's implementation (seaborn/sci-kit learn) of a heatmap, make it so that species will be in the same order each time to make comparison between different generations of matrices easier...