Add BirdNET-Gemodel 3.0

Author: Josef-Haupt

docs/birdnet.geo_models.v3_0.rst

@@ -0,0 +1,29 @@
+birdnet.geo\_models.v3\_0 package
+=================================
+
+Submodules
+----------
+
+birdnet.geo\_models.v3\_0.model module
+--------------------------------------
+
+.. automodule:: birdnet.geo_models.v3_0.model
+   :members:
+   :show-inheritance:
+   :undoc-members:
+
+birdnet.geo\_models.v3\_0.tf module
+------------------------------------
+
+.. automodule:: birdnet.geo_models.v3_0.tf
+   :members:
+   :show-inheritance:
+   :undoc-members:
+
+Module contents
+---------------
+
+.. automodule:: birdnet.geo_models.v3_0
+   :members:
+   :show-inheritance:
+   :undoc-members:

src/birdnet/geo/models/v3_0/__init__.py

@@ -0,0 +1 @@
+#

src/birdnet/geo/models/v3_0/model.py

@@ -0,0 +1,173 @@
+from __future__ import annotations
+
+from pathlib import Path
+from typing import Any, final
+
+from ordered_set import OrderedSet
+
+from birdnet.core.backends import (
+  BackendLoader,
+  VersionedGeoBackendProtocol,
+)
+from birdnet.geo.inference.prediction_result import GeoPredictionResult
+from birdnet.geo.inference.session import GeoPredictionSession
+from birdnet.geo.models.base import GeoModelBase
+from birdnet.globals import (
+  GEO_MODEL_VERSION_V3_0,
+  GEO_MODEL_VERSIONS,
+  MODEL_TYPE_GEO,
+  MODEL_TYPES,
+)
+from birdnet.utils.helper import validate_species_list
+
+
+class GeoDownloaderBaseV3_0:
+  AVAILABLE_LANGUAGES: OrderedSet[str] = OrderedSet(
+    (
+      "af",
+      "ar",
+      "cs",
+      "da",
+      "de",
+      "en_uk",
+      "en_us",
+      "es",
+      "fi",
+      "fr",
+      "hu",
+      "it",
+      "ja",
+      "ko",
+      "nl",
+      "no",
+      "pl",
+      "pt",
+      "ro",
+      "ru",
+      "sk",
+      "sl",
+      "sv",
+      "th",
+      "tr",
+      "uk",
+      "zh",
+    )
+  )
+
+
+class GeoModelV3_0(GeoModelBase):
+  def __init__(
+    self,
+    model_path: Path,
+    species_list: OrderedSet[str],
+    is_custom_model: bool,
+    backend_type: type[VersionedGeoBackendProtocol],
+    backend_kwargs: dict[str, Any],
+  ) -> None:
+    super().__init__(
+      model_path, species_list, is_custom_model, backend_type, backend_kwargs
+    )
+
+  @classmethod
+  def load(
+    cls,
+    model_path: Path,
+    species_list: OrderedSet[str],
+    backend_type: type[VersionedGeoBackendProtocol],
+    backend_kwargs: dict[str, Any],
+  ) -> GeoModelV3_0:
+    result = GeoModelV3_0(
+      model_path,
+      species_list,
+      is_custom_model=False,
+      backend_type=backend_type,
+      backend_kwargs=backend_kwargs,
+    )
+    return result
+
+  @classmethod
+  def load_custom(
+    cls,
+    model_path: Path,
+    species_list: Path,
+    backend_type: type[VersionedGeoBackendProtocol],
+    backend_kwargs: dict[str, Any],
+    check_validity: bool,
+  ) -> GeoModelV3_0:
+    assert model_path.exists()
+    assert species_list.is_file()
+
+    loaded_species_list = validate_species_list(species_list)
+
+    if check_validity:
+      n_species_in_model = BackendLoader.check_model_can_be_loaded(
+        model_path, backend_type, backend_kwargs
+      )
+
+      if n_species_in_model != len(loaded_species_list):
+        raise ValueError(
+          f"Model '{model_path.absolute()}' has {n_species_in_model} outputs, "
+          f"but species list '{species_list.absolute()}' "
+          f"has {len(loaded_species_list)} species!"
+        )
+
+    result = GeoModelV3_0(
+      model_path,
+      loaded_species_list,
+      is_custom_model=True,
+      backend_type=backend_type,
+      backend_kwargs=backend_kwargs,
+    )
+    return result
+
+  @classmethod
+  @final
+  def get_version(cls) -> GEO_MODEL_VERSIONS:
+    return GEO_MODEL_VERSION_V3_0
+
+  @classmethod
+  @final
+  def get_model_type(cls) -> MODEL_TYPES:
+    return MODEL_TYPE_GEO
+
+  def predict_session(
+    self,
+    /,
+    *,
+    min_confidence: float = 0.03,
+    half_precision: bool = False,
+    device: str = "CPU",
+  ) -> GeoPredictionSession:
+    return GeoPredictionSession(
+      species_list=self.species_list,
+      model_path=self.model_path,
+      model_is_custom=self.is_custom_model,
+      model_version=self.get_version(),
+      model_backend_type=self.backend_type,
+      model_backend_custom_kwargs=self.backend_kwargs,
+      min_confidence=min_confidence,
+      half_precision=half_precision,
+      device=device,
+    )
+
+  def predict(
+    self,
+    latitude: float,
+    longitude: float,
+    /,
+    *,
+    week: int | None = None,
+    min_confidence: float = 0.03,
+    half_precision: bool = False,
+    device: str = "CPU",
+  ) -> GeoPredictionResult:
+    with self.predict_session(
+      min_confidence=min_confidence,
+      half_precision=half_precision,
+      device=device,
+    ) as session:
+      return session.run(
+        latitude,
+        longitude,
+        week=week,
+      )