Skip to content

Detector

Bases: BaseComponent

Runs object detection on image tiles.

Requirements
  • tiles_path: Directory containing tiles to process
Produces
  • infer_gdf: GeoDataFrame with detected bounding boxes
  • Columns: geometry, object_id, tile_path, detector_score, detector_class
Source code in canopyrs/engine/components/detector.py 
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
class DetectorComponent(BaseComponent):
    """
    Runs object detection on image tiles.

    Requirements:
        - tiles_path: Directory containing tiles to process

    Produces:
        - infer_gdf: GeoDataFrame with detected bounding boxes
        - Columns: geometry, object_id, tile_path, detector_score, detector_class
    """

    name = 'detector'

    BASE_REQUIRES_STATE = {StateKey.TILES_PATH}
    BASE_REQUIRES_COLUMNS: Set[str] = set()

    BASE_PRODUCES_STATE = {StateKey.INFER_GDF, StateKey.INFER_COCO_PATH}
    BASE_PRODUCES_COLUMNS = {Col.GEOMETRY, Col.OBJECT_ID, Col.TILE_PATH, Col.DETECTOR_SCORE, Col.DETECTOR_CLASS}

    BASE_STATE_HINTS = {
        StateKey.TILES_PATH: (
            "Detector needs tiles to process. Add a tilerizer before detector."
        ),
    }

    BASE_COLUMN_HINTS: dict = {}

    def __init__(
        self,
        config: DetectorConfig,
        parent_output_path: str = None,
        component_id: int = None
    ):
        super().__init__(config, parent_output_path, component_id)

        # Store model class (instantiate in __call__ to avoid loading during validation)
        if config.model not in DETECTOR_REGISTRY:
            raise ValueError(f'Invalid detector model: {config.model}')
        self._model_class = DETECTOR_REGISTRY.get(config.model)

        # Set requirements
        self.requires_state = set(self.BASE_REQUIRES_STATE)
        self.requires_columns = set(self.BASE_REQUIRES_COLUMNS)
        self.produces_state = set(self.BASE_PRODUCES_STATE)
        self.produces_columns = set(self.BASE_PRODUCES_COLUMNS)

        # Set hints
        self.state_hints = dict(self.BASE_STATE_HINTS)
        self.column_hints = dict(self.BASE_COLUMN_HINTS)

    @classmethod
    def run_standalone(
        cls,
        config: DetectorConfig,
        tiles_path: str,
        output_path: str,
    ) -> 'DataState':
        """
        Run detector standalone on pre-tiled imagery.

        Args:
            config: Detector configuration
            tiles_path: Path to directory containing tiles
            output_path: Where to save outputs

        Returns:
            DataState with detection results (access .infer_gdf for the GeoDataFrame)

        Example:
            result = DetectorComponent.run_standalone(
                config=DetectorConfig(model='faster_rcnn_detectron2', ...),
                tiles_path='./tiles',
                output_path='./output',
            )
            print(result.infer_gdf)
        """
        from canopyrs.engine.pipeline import run_component
        return run_component(
            component=cls(config),
            output_path=output_path,
            tiles_path=tiles_path,
        )

    @validate_requirements
    def __call__(self, data_state: DataState) -> ComponentResult:
        """
        Run object detection on tiles.

        Returns flattened GDF. Pipeline handles merging and object_id assignment.
        """

        detector = self._model_class(self.config)

        # Create dataset from tiles
        infer_ds = UnlabeledRasterDataset(
            fold=None,
            root_path=data_state.tiles_path,
            transform=None
        )

        # Run inference
        tiles_paths, boxes, boxes_scores, classes = detector.infer(infer_ds, collate_fn_images)

        # Flatten outputs into GDF
        rows = []
        unique_id = 0
        for i, tile_path in enumerate(tiles_paths):
            for box_geom, score, cls_id in zip(boxes[i], boxes_scores[i], classes[i]):
                rows.append({
                    Col.GEOMETRY: box_geom,
                    Col.TILE_PATH: str(tile_path),
                    Col.DETECTOR_SCORE: score,
                    Col.DETECTOR_CLASS: cls_id,
                    Col.OBJECT_ID: unique_id,
                })
                unique_id += 1

        # Create GDF
        if not rows:
            gdf = gpd.GeoDataFrame(
                columns=list(self.produces_columns),
                crs=None
            )
        else:
            gdf = gpd.GeoDataFrame(rows, geometry=Col.GEOMETRY, crs=None)
            # Ensure geometry is valid
            gdf[Col.GEOMETRY] = gdf[Col.GEOMETRY].buffer(0)
            gdf = gdf[gdf.is_valid & ~gdf.is_empty]

        print(f"DetectorComponent: Generated {len(gdf)} detections.")

        return ComponentResult(
            gdf=gdf,
            produced_columns=self.produces_columns,
            objects_are_new=True,
            save_gpkg=True,
            gpkg_name_suffix="notaggregated",
            save_coco=True,
            coco_scores_column=Col.DETECTOR_SCORE,
            coco_categories_column=Col.DETECTOR_CLASS,
        )

__call__(data_state)

Run object detection on tiles.

Returns flattened GDF. Pipeline handles merging and object_id assignment.

Source code in canopyrs/engine/components/detector.py 
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
@validate_requirements
def __call__(self, data_state: DataState) -> ComponentResult:
    """
    Run object detection on tiles.

    Returns flattened GDF. Pipeline handles merging and object_id assignment.
    """

    detector = self._model_class(self.config)

    # Create dataset from tiles
    infer_ds = UnlabeledRasterDataset(
        fold=None,
        root_path=data_state.tiles_path,
        transform=None
    )

    # Run inference
    tiles_paths, boxes, boxes_scores, classes = detector.infer(infer_ds, collate_fn_images)

    # Flatten outputs into GDF
    rows = []
    unique_id = 0
    for i, tile_path in enumerate(tiles_paths):
        for box_geom, score, cls_id in zip(boxes[i], boxes_scores[i], classes[i]):
            rows.append({
                Col.GEOMETRY: box_geom,
                Col.TILE_PATH: str(tile_path),
                Col.DETECTOR_SCORE: score,
                Col.DETECTOR_CLASS: cls_id,
                Col.OBJECT_ID: unique_id,
            })
            unique_id += 1

    # Create GDF
    if not rows:
        gdf = gpd.GeoDataFrame(
            columns=list(self.produces_columns),
            crs=None
        )
    else:
        gdf = gpd.GeoDataFrame(rows, geometry=Col.GEOMETRY, crs=None)
        # Ensure geometry is valid
        gdf[Col.GEOMETRY] = gdf[Col.GEOMETRY].buffer(0)
        gdf = gdf[gdf.is_valid & ~gdf.is_empty]

    print(f"DetectorComponent: Generated {len(gdf)} detections.")

    return ComponentResult(
        gdf=gdf,
        produced_columns=self.produces_columns,
        objects_are_new=True,
        save_gpkg=True,
        gpkg_name_suffix="notaggregated",
        save_coco=True,
        coco_scores_column=Col.DETECTOR_SCORE,
        coco_categories_column=Col.DETECTOR_CLASS,
    )

run_standalone(config, tiles_path, output_path) classmethod

Run detector standalone on pre-tiled imagery.

Parameters:

Name Type Description Default
config DetectorConfig

Detector configuration

 required
tiles_path str

Path to directory containing tiles

 required
output_path str

Where to save outputs

 required

Returns:

Type Description
DataState

DataState with detection results (access .infer_gdf for the GeoDataFrame)
Example

result = DetectorComponent.run_standalone( config=DetectorConfig(model='faster_rcnn_detectron2', ...), tiles_path='./tiles', output_path='./output', ) print(result.infer_gdf)

Source code in canopyrs/engine/components/detector.py 
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
@classmethod
def run_standalone(
    cls,
    config: DetectorConfig,
    tiles_path: str,
    output_path: str,
) -> 'DataState':
    """
    Run detector standalone on pre-tiled imagery.

    Args:
        config: Detector configuration
        tiles_path: Path to directory containing tiles
        output_path: Where to save outputs

    Returns:
        DataState with detection results (access .infer_gdf for the GeoDataFrame)

    Example:
        result = DetectorComponent.run_standalone(
            config=DetectorConfig(model='faster_rcnn_detectron2', ...),
            tiles_path='./tiles',
            output_path='./output',
        )
        print(result.infer_gdf)
    """
    from canopyrs.engine.pipeline import run_component
    return run_component(
        component=cls(config),
        output_path=output_path,
        tiles_path=tiles_path,
    )