Merge remote-tracking branch 'origin/main' into focal-loss-autotune

Author: Josef-Haupt

.github/workflows/ci.yml

@@ -16,10 +16,12 @@ concurrency:
 jobs:
   running-tests:
     runs-on: ${{ matrix.os }}
+    env:
+      IS_GITHUB_RUNNER: "true"
     strategy:
       matrix:
         os: [ubuntu-latest, macos-latest, windows-latest]
-        python-version: ["3.11"]
+        python-version: ["3.11", "3.12"]
     steps:
       - uses: actions/checkout@v4
         with:
@@ -32,9 +34,7 @@ jobs:
       - name: Install dependencies
         run: |
           python -m pip install --upgrade pip
-          python -m pip install .[embeddings,train]
+          python -m pip install .[embeddings,train,tests]
       - name: Run tests
         run: |
-          python -m pip install .[tests]
-          python -m birdnet_analyzer.utils
           python -m pytest

.github/workflows/documentation.yml

@@ -22,7 +22,7 @@ jobs:
       - uses: actions/checkout@v4
       - uses: actions/setup-python@v5
         with:
-          python-version: '3.11'
+          python-version: '3.12'
       - name: Install dependencies
         run: |
           pip install .[docs]

.github/workflows/lint.yml

@@ -17,7 +17,7 @@ jobs:
       - name: Set up Python
         uses: actions/setup-python@v4
         with:
-          python-version: '3.11'
+          python-version: '3.12'
       - name: Install dependencies
         run: |
           python -m pip install --upgrade pip

.github/workflows/publish.yml

@@ -20,7 +20,7 @@ jobs:
       - name: Set up Python
         uses: actions/setup-python@v4
         with:
-          python-version: "3.11"
+          python-version: "3.12"
 
       - name: Install dependencies
         run: |

.github/workflows/test-publish.yml

@@ -21,7 +21,7 @@ jobs:
       - name: Set up Python
         uses: actions/setup-python@v4
         with:
-          python-version: "3.11"
+          python-version: "3.12"
 
       - name: Install dependencies
         run: |

.gitignore

@@ -145,8 +145,8 @@ celerybeat.pid
 *.sage.py
 
 # Environments
-.env
-.venv
+.env*
+.venv*
 env/
 venv/
 ENV/

README.md

@@ -9,7 +9,7 @@
 
 ![License](https://img.shields.io/github/license/birdnet-team/BirdNET-Analyzer)
 ![OS](https://badgen.net/badge/OS/Linux%2C%20Windows%2C%20macOS/blue)
-[![Python 3.11](https://img.shields.io/badge/python-3.11-blue.svg)](https://www.python.org/downloads/release/python-3110/)
+[![Python 3.12](https://img.shields.io/badge/python-3.12-blue.svg)](https://www.python.org/downloads/release/python-3120/)
 ![Species](https://badgen.net/badge/Species/6512/blue)
 ![Downloads](https://www-user.tu-chemnitz.de/~johau/birdnet_total_downloads_badge.php)
 

benchmark.sh

@@ -1,9 +1,18 @@
+import os
+import warnings
+
+from absl import logging
+
 from birdnet_analyzer.analyze import analyze
 from birdnet_analyzer.embeddings import embeddings
 from birdnet_analyzer.search import search
 from birdnet_analyzer.segments import segments
 from birdnet_analyzer.species import species
 from birdnet_analyzer.train import train
 
-__version__ = "2.0.0"
+__version__ = "2.2.0"
 __all__ = ["analyze", "embeddings", "search", "segments", "species", "train"]
+os.environ["TF_CPP_MIN_LOG_LEVEL"] = "3"
+
+logging.set_verbosity(logging.ERROR)
+warnings.filterwarnings("ignore")

birdnet_analyzer/__init__.py

@@ -27,6 +27,7 @@ def analyze(
     threads: int = 8,
     locale: str = "en",
     additional_columns: list[str] | None = None,
+    use_perch: bool = False,
 ):
     """
     Analyzes audio files for bird species detection using the BirdNET-Analyzer.
@@ -55,6 +56,7 @@ def analyze(
         threads (int, optional): Number of CPU threads to use for analysis. Defaults to 8.
         locale (str, optional): Locale for species names and output. Defaults to "en".
         additional_columns (list[str] | None, optional): Additional columns to include in the output. Defaults to None.
+        use_perch (bool, optional): Whether to use the Perch model for analysis. Defaults to False.
     Returns:
         None
     Raises:
@@ -69,9 +71,6 @@ def analyze(
     import birdnet_analyzer.config as cfg
     from birdnet_analyzer.analyze.utils import analyze_file, save_analysis_params
     from birdnet_analyzer.analyze.utils import combine_results as combine
-    from birdnet_analyzer.utils import ensure_model_exists
-
-    ensure_model_exists()
 
     flist = _set_params(
         audio_input=audio_input,
@@ -98,6 +97,7 @@ def analyze(
         threads=threads,
         labels_file=cfg.LABELS_FILE,
         additional_columns=additional_columns,
+        use_perch=use_perch,
     )
 
     print(f"Found {len(cfg.FILE_LIST)} files to analyze")
@@ -154,29 +154,37 @@ def _set_params(
     threads,
     labels_file=None,
     additional_columns=None,
+    use_perch=False,
 ):
     import birdnet_analyzer.config as cfg
     from birdnet_analyzer.analyze.utils import load_codes
     from birdnet_analyzer.species.utils import get_species_list
-    from birdnet_analyzer.utils import collect_audio_files, read_lines
+    from birdnet_analyzer.utils import collect_audio_files, ensure_model_exists, read_lines
+
+    ensure_model_exists(check_perch=use_perch)
 
     if not isinstance(overlap, int | float):
         raise ValueError("Overlap must be a numeric value.")
 
     if overlap < 0:
         raise ValueError("Overlap must be a non-negative value.")
 
-    if overlap >= cfg.SIG_LENGTH:
-        raise ValueError(f"Overlap must be less than {cfg.SIG_LENGTH} seconds.")
+    if not use_perch and overlap > 2.9:
+        raise ValueError("Overlap must be less than or equal to 2.9 seconds for BirdNET model.")
+
+    if use_perch and overlap > 4.9:
+        raise ValueError("Overlap must be less than or equal to 4.9 seconds for Perch model.")
 
     if not isinstance(audio_speed, int | float):
         raise ValueError("Audio speed must be a numeric value.")
 
     if audio_speed <= 0:
         raise ValueError("Audio speed must be a positive value.")
 
+    if use_perch and sensitivity != 1.0:
+        print("Warning: Sensitivity setting is ignored when using the Perch model.")
+
     cfg.CODES = load_codes()
-    cfg.LABELS = read_lines(labels_file if labels_file else cfg.LABELS_FILE)
     cfg.SKIP_EXISTING_RESULTS = skip_existing_results
     cfg.LOCATION_FILTER_THRESHOLD = sf_thresh
     cfg.TOP_N = top_n
@@ -192,6 +200,10 @@ def _set_params(
     cfg.COMBINE_RESULTS = combine_results
     cfg.BATCH_SIZE = bs
     cfg.ADDITIONAL_COLUMNS = additional_columns
+    cfg.USE_PERCH = use_perch
+
+    if cfg.USE_PERCH and custom_classifier:
+        raise ValueError("Selected custom classifier and Perch model, please select only one.")
 
     if not output:
         if os.path.isfile(cfg.INPUT_PATH):
@@ -213,47 +225,66 @@ def _set_params(
         cfg.CPU_THREADS = 1
         cfg.TFLITE_THREADS = threads
 
-    if custom_classifier is not None:
+    if cfg.USE_PERCH:
+        cfg.MODEL_PATH = cfg.PERCH_V2_MODEL_PATH
+        cfg.LABELS_FILE = cfg.PERCH_LABELS_FILE
+        cfg.SAMPLE_RATE = cfg.PERCH_SAMPLE_RATE
+        cfg.SIG_LENGTH = cfg.PERCH_SIG_LENGTH
+        cfg.LABELS = read_lines(cfg.PERCH_LABELS_FILE)
+        cfg.LABELS = cfg.LABELS[1:]  # it's a csv with header
+    else:
+        cfg.MODEL_PATH = cfg.BIRDNET_MODEL_PATH
+        cfg.LABELS_FILE = cfg.BIRDNET_LABELS_FILE
+        cfg.SAMPLE_RATE = cfg.BIRDNET_SAMPLE_RATE
+        cfg.SIG_LENGTH = cfg.BIRDNET_SIG_LENGTH
+        cfg.LABELS = read_lines(labels_file if labels_file else cfg.LABELS_FILE)
+
+    if overlap >= cfg.SIG_LENGTH:
+        raise ValueError(f"Overlap must be less than {cfg.SIG_LENGTH} seconds.")
+
+    # Custom classifier trained with the Analyzer, not arbitrary models, meaning; A a tflite model or B a raven model
+    if custom_classifier is None:
+        # TODO: does species list even make sense with Perch?
+        cfg.LATITUDE, cfg.LONGITUDE, cfg.WEEK = lat, lon, week
+        cfg.CUSTOM_CLASSIFIER = None
+
+        if cfg.LATITUDE == -1 and cfg.LONGITUDE == -1:
+            if not slist:
+                cfg.SPECIES_LIST_FILE = None
+            else:
+                cfg.SPECIES_LIST_FILE = slist
+
+                if os.path.isdir(cfg.SPECIES_LIST_FILE):
+                    cfg.SPECIES_LIST_FILE = os.path.join(cfg.SPECIES_LIST_FILE, "species_list.txt")
+
+            cfg.SPECIES_LIST = read_lines(cfg.SPECIES_LIST_FILE, trim=True, fail_on_blank_lines=True)
+        else:
+            cfg.SPECIES_LIST_FILE = None
+            cfg.SPECIES_LIST = get_species_list(cfg.LATITUDE, cfg.LONGITUDE, cfg.WEEK, cfg.LOCATION_FILTER_THRESHOLD)
+    else:
         cfg.CUSTOM_CLASSIFIER = custom_classifier  # we treat this as absolute path, so no need to join with dirname
 
         if custom_classifier.endswith(".tflite"):
             cfg.LABELS_FILE = custom_classifier.replace(".tflite", "_Labels.txt")  # same for labels file
 
-            if not os.path.isfile(cfg.LABELS_FILE): # if the label file is not found, an old birdnet model might be used
+            if not os.path.isfile(cfg.LABELS_FILE):  # if the label file is not found, an old birdnet model might be used
                 cfg.LABELS_FILE = custom_classifier.replace("Model_FP32.tflite", "Labels.txt")
 
-            if not os.path.isfile(cfg.LABELS_FILE): # if the label file is still not found, dont use labels
+            if not os.path.isfile(cfg.LABELS_FILE):  # if the label file is still not found, dont use labels
                 cfg.LABELS_FILE = None
                 cfg.LABELS = None
             else:
-                cfg.LABELS = read_lines(cfg.LABELS_FILE)
+                cfg.LABELS = read_lines(cfg.LABELS_FILE, fail_on_blank_lines=True)
         else:
             cfg.APPLY_SIGMOID = False
             # our output format
             cfg.LABELS_FILE = os.path.join(custom_classifier, "labels", "label_names.csv")
 
             if not os.path.isfile(cfg.LABELS_FILE):
                 cfg.LABELS_FILE = os.path.join(custom_classifier, "assets", "label.csv")
-                cfg.LABELS = read_lines(cfg.LABELS_FILE)
-            else:
-                cfg.LABELS = [line.split(",")[1] for line in read_lines(cfg.LABELS_FILE)]
-    else:
-        cfg.LATITUDE, cfg.LONGITUDE, cfg.WEEK = lat, lon, week
-        cfg.CUSTOM_CLASSIFIER = None
-
-        if cfg.LATITUDE == -1 and cfg.LONGITUDE == -1:
-            if not slist:
-                cfg.SPECIES_LIST_FILE = None
+                cfg.LABELS = read_lines(cfg.LABELS_FILE, fail_on_blank_lines=True)
             else:
-                cfg.SPECIES_LIST_FILE = slist
-
-                if os.path.isdir(cfg.SPECIES_LIST_FILE):
-                    cfg.SPECIES_LIST_FILE = os.path.join(cfg.SPECIES_LIST_FILE, "species_list.txt")
-
-            cfg.SPECIES_LIST = read_lines(cfg.SPECIES_LIST_FILE)
-        else:
-            cfg.SPECIES_LIST_FILE = None
-            cfg.SPECIES_LIST = get_species_list(cfg.LATITUDE, cfg.LONGITUDE, cfg.WEEK, cfg.LOCATION_FILTER_THRESHOLD)
+                cfg.LABELS = [line.split(",")[1] for line in read_lines(cfg.LABELS_FILE, fail_on_blank_lines=True)]
 
     if cfg.LABELS_FILE:
         lfile = os.path.join(cfg.TRANSLATED_LABELS_PATH, os.path.basename(cfg.LABELS_FILE).replace(".txt", f"_{locale}.txt"))