geoai-py 0.9.0__py2.py3-none-any.whl → 0.9.2__py2.py3-none-any.whl

geoai/__init__.py

@@ -2,7 +2,7 @@
 
 __author__ = """Qiusheng Wu"""
 __email__ = "giswqs@gmail.com"
-__version__ = "0.9.0"
+__version__ = "0.9.2"
 
 
 import os

geoai/change_detection.py

@@ -1,13 +1,16 @@
 """Change detection module for remote sensing imagery using torchange."""
 
 import os
+from typing import Any, Dict, List, Optional, Tuple, Union
+
+import cv2
+import matplotlib.pyplot as plt
 import numpy as np
 import rasterio
 from rasterio.windows import from_bounds
 from skimage.transform import resize
-import matplotlib.pyplot as plt
-import cv2
 from torchange.models.segment_any_change import AnyChange, show_change_masks
+
 from .utils import download_file
 
 
@@ -47,15 +50,15 @@ class ChangeDetection:
 
     def set_hyperparameters(
         self,
-        change_confidence_threshold=155,
-        auto_threshold=False,
-        use_normalized_feature=True,
-        area_thresh=0.8,
-        match_hist=False,
-        object_sim_thresh=60,
-        bitemporal_match=True,
-        **kwargs,
-    ):
+        change_confidence_threshold: int = 155,
+        auto_threshold: bool = False,
+        use_normalized_feature: bool = True,
+        area_thresh: float = 0.8,
+        match_hist: bool = False,
+        object_sim_thresh: int = 60,
+        bitemporal_match: bool = True,
+        **kwargs: Any,
+    ) -> None:
         """
         Set hyperparameters for the change detection model.
 
@@ -83,16 +86,16 @@ class ChangeDetection:
 
     def set_mask_generator_params(
         self,
-        points_per_side=32,
+        points_per_side: int = 32,
         points_per_batch: int = 64,
         pred_iou_thresh: float = 0.5,
         stability_score_thresh: float = 0.95,
         stability_score_offset: float = 1.0,
         box_nms_thresh: float = 0.7,
-        point_grids=None,
+        point_grids: Optional[List] = None,
         min_mask_region_area: int = 0,
-        **kwargs,
-    ):
+        **kwargs: Any,
+    ) -> None:
         """
         Set mask generator parameters.
 
@@ -203,17 +206,17 @@ class ChangeDetection:
 
     def detect_changes(
         self,
-        image1_path,
-        image2_path,
-        output_path=None,
-        target_size=1024,
-        return_results=True,
-        export_probability=False,
-        probability_output_path=None,
-        export_instance_masks=False,
-        instance_masks_output_path=None,
-        return_detailed_results=False,
-    ):
+        image1_path: str,
+        image2_path: str,
+        output_path: Optional[str] = None,
+        target_size: int = 1024,
+        return_results: bool = True,
+        export_probability: bool = False,
+        probability_output_path: Optional[str] = None,
+        export_instance_masks: bool = False,
+        instance_masks_output_path: Optional[str] = None,
+        return_detailed_results: bool = False,
+    ) -> Union[Tuple[Any, np.ndarray, np.ndarray], Dict[str, Any], None]:
         """
         Detect changes between two GeoTIFF images with instance segmentation.
 
@@ -530,7 +533,9 @@ class ChangeDetection:
         ) as dst:
             dst.write(prob_final.astype(np.float32), 1)
 
-    def visualize_changes(self, image1_path, image2_path, figsize=(15, 5)):
+    def visualize_changes(
+        self, image1_path: str, image2_path: str, figsize: Tuple[int, int] = (15, 5)
+    ) -> plt.Figure:
         """
         Visualize change detection results.
 
@@ -1516,14 +1521,16 @@ Areas:
         return results
 
 
-def download_checkpoint(model_type="vit_h", checkpoint_dir=None):
+def download_checkpoint(
+    model_type: str = "vit_h", checkpoint_dir: Optional[str] = None
+) -> str:
     """Download the SAM model checkpoint.
 
     Args:
         model_type (str, optional): The model type. Can be one of ['vit_h', 'vit_l', 'vit_b'].
             Defaults to 'vit_h'. See https://bit.ly/3VrpxUh for more details.
         checkpoint_dir (str, optional): The checkpoint_dir directory. Defaults to None,
-        "~/.cache/torch/hub/checkpoints".
+            which uses "~/.cache/torch/hub/checkpoints".
     """
 
     model_types = {

geoai/classify.py

@@ -1,50 +1,51 @@
 """The module for training semantic segmentation models for classifying remote sensing imagery."""
 
 import os
+from typing import Any, Dict, List, Optional, Union
 
 import numpy as np
 
 
 def train_classifier(
-    image_root,
-    label_root,
-    output_dir="output",
-    in_channels=4,
-    num_classes=14,
-    epochs=20,
-    img_size=256,
-    batch_size=8,
-    sample_size=500,
-    model="unet",
-    backbone="resnet50",
-    weights=True,
-    num_filters=3,
-    loss="ce",
-    class_weights=None,
-    ignore_index=None,
-    lr=0.001,
-    patience=10,
-    freeze_backbone=False,
-    freeze_decoder=False,
-    transforms=None,
-    use_augmentation=False,
-    seed=42,
-    train_val_test_split=(0.6, 0.2, 0.2),
-    accelerator="auto",
-    devices="auto",
-    logger=None,
-    callbacks=None,
-    log_every_n_steps=10,
-    use_distributed_sampler=False,
-    monitor_metric="val_loss",
-    mode="min",
-    save_top_k=1,
-    save_last=True,
-    checkpoint_filename="best_model",
-    checkpoint_path=None,
-    every_n_epochs=1,
-    **kwargs,
-):
+    image_root: str,
+    label_root: str,
+    output_dir: str = "output",
+    in_channels: int = 4,
+    num_classes: int = 14,
+    epochs: int = 20,
+    img_size: int = 256,
+    batch_size: int = 8,
+    sample_size: int = 500,
+    model: str = "unet",
+    backbone: str = "resnet50",
+    weights: bool = True,
+    num_filters: int = 3,
+    loss: str = "ce",
+    class_weights: Optional[List[float]] = None,
+    ignore_index: Optional[int] = None,
+    lr: float = 0.001,
+    patience: int = 10,
+    freeze_backbone: bool = False,
+    freeze_decoder: bool = False,
+    transforms: Optional[Any] = None,
+    use_augmentation: bool = False,
+    seed: int = 42,
+    train_val_test_split: tuple = (0.6, 0.2, 0.2),
+    accelerator: str = "auto",
+    devices: str = "auto",
+    logger: Optional[Any] = None,
+    callbacks: Optional[List[Any]] = None,
+    log_every_n_steps: int = 10,
+    use_distributed_sampler: bool = False,
+    monitor_metric: str = "val_loss",
+    mode: str = "min",
+    save_top_k: int = 1,
+    save_last: bool = True,
+    checkpoint_filename: str = "best_model",
+    checkpoint_path: Optional[str] = None,
+    every_n_epochs: int = 1,
+    **kwargs: Any,
+) -> Any:
     """Train a semantic segmentation model on geospatial imagery.
 
     This function sets up datasets, model, trainer, and executes the training process
@@ -584,15 +585,15 @@ def _classify_image(
 
 
 def classify_image(
-    image_path,
-    model_path,
-    output_path=None,
-    chip_size=1024,
-    overlap=256,
-    batch_size=4,
-    colormap=None,
-    **kwargs,
-):
+    image_path: str,
+    model_path: str,
+    output_path: Optional[str] = None,
+    chip_size: int = 1024,
+    overlap: int = 256,
+    batch_size: int = 4,
+    colormap: Optional[Dict] = None,
+    **kwargs: Any,
+) -> str:
     """
     Classify a geospatial image using a trained semantic segmentation model.
 
@@ -826,15 +827,15 @@ def classify_image(
 
 
 def classify_images(
-    image_paths,
-    model_path,
-    output_dir=None,
-    chip_size=1024,
-    batch_size=4,
-    colormap=None,
-    file_extension=".tif",
-    **kwargs,
-):
+    image_paths: Union[str, List[str]],
+    model_path: str,
+    output_dir: Optional[str] = None,
+    chip_size: int = 1024,
+    batch_size: int = 4,
+    colormap: Optional[Dict] = None,
+    file_extension: str = ".tif",
+    **kwargs: Any,
+) -> List[str]:
     """
     Classify multiple geospatial images using a trained semantic segmentation model.