custom classifier AND species list

[mo-the-floh]

hello, I'm new to BirdNet and also github, so forgive me if I don't know any "rules" here and am unknowingly breaking them... ;)

I was wondering if it's possible to run a custom classifier, in my case just an older model (V2.1), AND including a species list, because I'm looking for just some specific bird species (which are included in the general birdnet species list). currently using the GUI (on mac).
How would I go about that?

Thanks in advance,
Mo

[Josef-Haupt]

You are not breaking any rules, it's a valid question.
We currently do not support an further filtering of the species for custom classifiers because the number of species for a cc is managable, most of the time. Exception being older BirdNET models and cc 's trained with append mode.
We'll add it to our backlog and implement it in a future version.

But you can always do your own post processing, simply lower the threshold, so that all or most species will be in the result table for each segment even if the confidence is low and manually filter the resulting table based on your own list of species list.

Here is an example script, I used in the past:

"""
Basic utility script to filter a Raven Selection table by values in a specified column.
"""

def filter_table(fpath: str, filter_col: str, filter_values: list[str]) -> None:
    """
    Filters a CSV file based on a specific column and value.

    Args:
        fpath (str): Path to the CSV file.
        filter_col (str): Column name to filter by.
        filter_value (str): Value to filter for in the specified column.
    """
    import pandas as pd

    # Read the CSV file
    df = pd.read_csv(fpath, sep='\t')

    # Filter the DataFrame
    filtered_df = df[df[filter_col].isin(filter_values)]

    # Save the filtered DataFrame to a new CSV file
    filtered_df.to_csv('filtered.txt', index=False, sep='\t')


path = r"C:\user\ZenodoData\Testdata2\S19_20180214_080003_resampled.BirdNET.selection.table.txt"

filter_table(path, "Common Name", ["noise", "gibbon.female", "hornbill.rhino"])