birdnetlib 0.5.0

A python api for BirdNET-Lite and BirdNET-Analyzer

birdnetlib

A python api for BirdNET-Analyzer and BirdNET-Lite

Installation

birdnetlib requires Python 3.7+ and prior installation of Tensorflow Lite, librosa and ffmpeg. See BirdNET-Analyzer for more details on installing the Tensorflow-related dependencies.

pip install birdnetlib

Documentation

birdnetlib provides a common interface for BirdNET-Analyzer and BirdNET-Lite.

Using BirdNET-Analyzer

To use the newer BirdNET-Analyzer model, use the Analyzer class.

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(year=2022, month=5, day=10), # use date or week_48
    min_conf=0.25,
)
recording.analyze()
print(recording.detections)

Using a custom classifier with BirdNET-Analyzer

To use the a model trained with BirdNET-Analyzer, use the Analyzer class.

from birdnetlib import Recording
from birdnetlib.analyzer import Analyzer

# Load and initialize BirdNET-Analyzer with your own model/labels.

custom_model_path = "custom_classifiers/trogoniformes.tflite"
custom_labels_path = "custom_classifiers/trogoniformes.txt"

analyzer = Analyzer(
    classifier_labels_path=custom_labels_path, classifier_model_path=custom_model_path
)

recording = Recording(
    analyzer,
    "sample.mp3",
    min_conf=0.25,
)
recording.analyze()
print(recording.detections)

Using BirdNET-Lite

To use the BirdNET-Lite model, use the LiteAnalyzer class.

from birdnetlib import Recording
from birdnetlib.analyzer_lite import LiteAnalyzer
from datetime import datetime

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

recording = Recording(
    analyzer,
    "sample.mp3",
    lat=35.4244,
    lon=-120.7463,
    date=datetime(year=2022, month=5, day=10), # use date or week_48
    min_conf=0.25,
)
recording.analyze()
print(recording.detections) # Returns list of detections.

recording.detections contains a list of detected species, along with time ranges and confidence value.

[{'common_name': 'House Finch',
  'confidence': 0.5744,
  'end_time': 12.0,
  'scientific_name': 'Haemorhous mexicanus',
  'start_time': 9.0},
 {'common_name': 'House Finch',
  'confidence': 0.4496,
  'end_time': 15.0,
  'scientific_name': 'Haemorhous mexicanus',
  'start_time': 12.0}]

Utility classes

DirectoryAnalyzer

DirectoryAnalyzer can process a directory and analyze contained files.

def on_analyze_complete(recording):
    print(recording.path)
    pprint(recording.detections)

directory = DirectoryAnalyzer(
    "/Birds/mp3_dir",
    patterns=["*.mp3", "*.wav"]
)
directory.on_analyze_complete = on_analyze_complete
directory.process()

See the full example for analyzer options and error handling callbacks.

DirectoryMultiProcessingAnalyzer

DirectoryMultiProcessingAnalyzer can process a directory and analyze contained files, using multiple processes asynchronously.

def on_analyze_directory_complete(recordings):
    for recording in recordings:
        pprint(recording.detections)

directory = "."
batch = DirectoryMultiProcessingAnalyzer(
    "/Birds/mp3_dir",
    patterns=["*.mp3", "*.wav"]
)

batch.on_analyze_directory_complete = on_analyze_directory_complete
batch.process()

See the full example for analyzer options and error handling callbacks.

DirectoryWatcher

DirectoryWatcher can watch a directory and analyze new files as they are created.

def on_analyze_complete(recording):
    print(recording.path)
    pprint(recording.detections)

watcher = DirectoryWatcher("/Birds/mp3_dir")
watcher.on_analyze_complete = on_analyze_complete
watcher.watch()

See the full example for analyzer options and error handling callbacks.

SpeciesList

SpeciesList uses BirdNET-Analyzer to predict species lists from location and date.

species = SpeciesList()
species_list = species.return_list(
    lon=-120.7463, lat=35.4244, date=datetime(year=2022, month=5, day=10)
)
print(species_list)
# [{'scientific_name': 'Haemorhous mexicanus', 'common_name': 'House Finch', 'threshold': 0.8916686}, ...]

Additional examples

• Watch a directory for new files, then analyze with both analyzer models as files are saved
• Watch a directory for new files, and apply datetimes by parsing file names (eg 2022-08-15-birdnet-21:05:52.wav) prior to analyzing This example can also be used to modify lat/lon, min_conf, etc., based on file name prior to analyzing.
• Limit detections to certain species by passing a predefined species list to the analyzer Useful when searching for a particular set of bird detections.
• Extract detections as audio file samples and/or spectrograms Supports audio extractions as .flac, .wav and .mp3. Spectrograms exported as .png, .jpg, or other matplotlib.pyplot supported formats. Can be filtered to only extract files above a separate minimum confidence value.

About BirdNET-Lite and BirdNET-Analyzer

birdnetlib uses models provided by BirdNET-Lite and BirdNET-Analyzer under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International Public License.

BirdNET-Lite and BirdNET-Analyzer were developed by the K. Lisa Yang Center for Conservation Bioacoustics at the Cornell Lab of Ornithology.

For more information on BirdNET analyzers, please see the project repositories below:

BirdNET-Analyzer

BirdNET-Lite

birdnetlib is not associated with BirdNET-Lite, BirdNET-Analyzer or the K. Lisa Yang Center for Conservation Bioacoustics.

About birdnetlib

birdnetlib is maintained by Joe Weiss. Contributions are welcome.

Project Goals

• Establish a unified API for interacting with Tensorflow-based BirdNET analyzers
• Enable python-based test cases for BirdNET analyzers
• Make it easier to use BirdNET in python-based projects
• Make it easier to migrate to new BirdNET versions/models as they become available