Skip to content

Conversation

@kgarwoodsdzwa
Copy link
Member

script to compare specifically how birdnet well does on different buowset versions

one script to convert the birdnet results for buowset into
a reusable pkl for the other script who will be able to assess
against the ground truth
now have a script that aligns the birdnet result along with
the buowset labels in one dataframe so we can easily compare the
2 with the next function i have to add. the data prep is done
just need to add in the function to compare the results
obtained a confusion matrix for birdnet on buowset. super high
precision, super low accuracy. doesnt seem to get false positives
but has mannyyyy false negatives. need to add docstrings and a way
to put these metrics of birdnet on cometml
this code allows you to obtain the performance metrics for
birdnet against a ground truth dataset (specifically buowset)
as a binary model, and also allows you to specify a certain
vocalization class to isolate the performance metrics for. it
also allows you to post these results to comet
there were some pylint and flake8 errors. This could be turned into
something in the whoot package itself but maybe should happen as
a refactor after merging to dev because this branch is behind dev
and doesn't have the __init__ stuff.
Copy link
Collaborator

@Sean1572 Sean1572 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Still need to finsh review

@kgarwoodsdzwa kgarwoodsdzwa requested a review from Sean1572 October 8, 2025 21:40
row_lower['MANUAL ID*'] = row['MANUAL ID*'].str.lower()

if row_lower['MANUAL ID*'] == 'cluck':
chunks_df.loc[start_chunk:end_chunk, 'Label'] = '0'
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is this intended? This is going to assign multiple rows to a given label

image

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants