What is the best way to filter out no bird species in the detection?

[aaron-dsouza]

I have used this very useful library to detect bird species from sounds in my backyard. However, it also detects human voices, engines and barking dogs. What is the best way to filter out detections of voices that are not bird or avian species?

[joeweiss]

Hi @aaron-dsouza. There isn't a built-in method for doing this, but you can pass your lat/lon and the current date to Recording like this:

from birdnetlib import Recording
from birdnetlib.analyzer import Analyzer
from datetime import datetime

# Load and initialize the BirdNET-Analyzer models.
analyzer = Analyzer()

recording = Recording(
    analyzer,
    "sample.mp3",
    lat=35.4244,
    lon=-120.7463,
    date=datetime.now().date(),
    min_conf=0.25,
)
recording.analyze()
print(recording.detections)

That will limit your detections to avian species that are expected for your current time and location.

The other alternative is to filter your recording.detections out post-analysis with a list comprehension like this:

filtered_items = ["Engine", "Human non-vocal", "Human vocal"] # Add as many as needed here ...
filtered_detections = [item for item in recording.detections if item["common_name"] not in filtered_items]

Hope that helps.