Refactored model.py to support image and LiDAR detection models.

detectionmetrics/models/model.py

@@ -12,10 +12,10 @@
 import detectionmetrics.utils.io as uio
 
 
-class SegmentationModel(ABC):
-    """Parent segmentation model class
+class PerceptionModel(ABC):
+    """Base class for all vision perception models (e.g., segmentation, detection).
 
-    :param model: Segmentation model object
+    :param model: Model object
     :type model: Any
     :param model_type: Model type (e.g. scripted, compiled, etc.)
     :type model_type: str
@@ -52,41 +52,14 @@ def __init__(
 
     @abstractmethod
     def inference(
-        self, points: Union[np.ndarray, Image.Image]
-    ) -> Union[np.ndarray, Image.Image]:
-        """Perform inference for a single image or point cloud
-
-        :param image: Either a numpy array (LiDAR point cloud) or a PIL image
-        :type image: Union[np.ndarray, Image.Image]
-        :return: Segmenation result as a point cloud or image with label indices
-        :rtype: Union[np.ndarray, Image.Image]
-        """
+        self, data: Union[np.ndarray, Image.Image]
+    ) -> Union[np.ndarray, Image.Image, dict]:
+        """Perform inference for a single image or point cloud."""
         raise NotImplementedError
 
     @abstractmethod
-    def eval(
-        self,
-        dataset: dm_dataset.SegmentationDataset,
-        split: str | List[str] = "test",
-        ontology_translation: Optional[str] = None,
-        predictions_outdir: Optional[str] = None,
-        results_per_sample: bool = False,
-    ) -> pd.DataFrame:
-        """Perform evaluation for an image segmentation dataset
-
-        :param dataset: Segmentation dataset for which the evaluation will be performed
-        :type dataset: ImageSegmentationDataset
-        :param split: Split or splits to be used from the dataset, defaults to "test"
-        :type split: str | List[str], optional
-        :param ontology_translation: JSON file containing translation between dataset and model output ontologies
-        :type ontology_translation: str, optional
-        :param predictions_outdir: Directory to save predictions per sample, defaults to None. If None, predictions are not saved.
-        :type predictions_outdir: Optional[str], optional
-        :param results_per_sample: Whether to store results per sample or not, defaults to False. If True, predictions_outdir must be provided.
-        :type results_per_sample: bool, optional
-        :return: DataFrame containing evaluation results
-        :rtype: pd.DataFrame
-        """
+    def eval(self, *args, **kwargs) -> pd.DataFrame:
+        """Evaluate the model on the given dataset."""
         raise NotImplementedError
 
     @abstractmethod
@@ -124,6 +97,71 @@ def get_lut_ontology(
         return lut_ontology
 
 
+class SegmentationModel(PerceptionModel):
+    """Parent segmentation model class
+
+    :param model: Segmentation model object
+    :type model: Any
+    :param model_type: Model type (e.g. scripted, compiled, etc.)
+    :type model_type: str
+    :param model_cfg: JSON file containing model configuration
+    :type model_cfg: str
+    :param ontology_fname: JSON file containing model output ontology
+    :type ontology_fname: str
+    :param model_fname: Model file or directory, defaults to None
+    :type model_fname: Optional[str], optional
+    """
+
+    def __init__(
+        self,
+        model: Any,
+        model_type: str,
+        model_cfg: str,
+        ontology_fname: str,
+        model_fname: Optional[str] = None,
+    ):
+        super().__init__(model, model_type, model_cfg, ontology_fname, model_fname)
+
+    @abstractmethod
+    def inference(
+        self, points: Union[np.ndarray, Image.Image]
+    ) -> Union[np.ndarray, Image.Image]:
+        """Perform inference for a single image or point cloud
+
+        :param image: Either a numpy array (LiDAR point cloud) or a PIL image
+        :type image: Union[np.ndarray, Image.Image]
+        :return: Segmenation result as a point cloud or image with label indices
+        :rtype: Union[np.ndarray, Image.Image]
+        """
+        raise NotImplementedError
+
+    @abstractmethod
+    def eval(
+        self,
+        dataset: dm_dataset.SegmentationDataset,
+        split: str | List[str] = "test",
+        ontology_translation: Optional[str] = None,
+        predictions_outdir: Optional[str] = None,
+        results_per_sample: bool = False,
+    ) -> pd.DataFrame:
+        """Perform evaluation for an image segmentation dataset
+
+        :param dataset: Segmentation dataset for which the evaluation will be performed
+        :type dataset: ImageSegmentationDataset
+        :param split: Split or splits to be used from the dataset, defaults to "test"
+        :type split: str | List[str], optional
+        :param ontology_translation: JSON file containing translation between dataset and model output ontologies
+        :type ontology_translation: str, optional
+        :param predictions_outdir: Directory to save predictions per sample, defaults to None. If None, predictions are not saved.
+        :type predictions_outdir: Optional[str], optional
+        :param results_per_sample: Whether to store results per sample or not, defaults to False. If True, predictions_outdir must be provided.
+        :type results_per_sample: bool, optional
+        :return: DataFrame containing evaluation results
+        :rtype: pd.DataFrame
+        """
+        raise NotImplementedError
+
+
 class ImageSegmentationModel(SegmentationModel):
     """Parent image segmentation model class
 
@@ -248,3 +286,191 @@ def eval(
         :rtype: pd.DataFrame
         """
         raise NotImplementedError
+
+class DetectionModel(PerceptionModel):
+    """Parent detection model class
+
+    :param model: Detection model object
+    :type model: Any
+    :param model_type: Model type (e.g. scripted, compiled, etc.)
+    :type model_type: str
+    :param model_cfg: JSON file containing model configuration
+    :type model_cfg: str
+    :param ontology_fname: JSON file containing model output ontology
+    :type ontology_fname: str
+    :param model_fname: Model file or directory, defaults to None
+    :type model_fname: Optional[str], optional
+    """
+
+    def __init__(
+        self,
+        model: Any,
+        model_type: str,
+        model_cfg: str,
+        ontology_fname: str,
+        model_fname: Optional[str] = None,
+    ):
+        super().__init__(model, model_type, model_cfg, ontology_fname, model_fname)
+
+    @abstractmethod
+    def inference(
+        self, data: Union[np.ndarray, Image.Image]
+    ) -> List[dict]:
+        """Perform inference for a single input (image or point cloud)
+
+        :param data: Input image or LiDAR point cloud
+        :type data: Union[np.ndarray, Image.Image]
+        :return: List of detection results (each a dict with bbox, confidence, class)
+        :rtype: List[dict]
+        """
+        raise NotImplementedError
+
+    @abstractmethod
+    def eval(
+        self,
+        dataset: dm_dataset.DetectionDataset,
+        split: Union[str, List[str]] = "test",
+        ontology_translation: Optional[str] = None,
+        predictions_outdir: Optional[str] = None,
+        results_per_sample: bool = False,
+    ) -> pd.DataFrame:
+        """Perform evaluation for a detection dataset
+
+        :param dataset: Detection dataset for which evaluation will be performed
+        :type dataset: ImageDetecctionDataset
+        :param split: Split(s) to use, defaults to "test"
+        :type split: Union[str, List[str]]
+        :param ontology_translation: JSON file containing translation between dataset and model output ontologies
+        :type ontology_translation: Optional[str]
+        :param predictions_outdir: Directory to save predictions per sample, defaults to None. If None, predictions are not saved.
+        :type predictions_outdir: Optional[str]
+        :param results_per_sample: Whether to store results per sample or not, defaults to False. If True, predictions_outdir must be provided.
+        :type results_per_sample: bool
+        :return: DataFrame containing evaluation metrics
+        :rtype: pd.DataFrame
+        """
+        raise NotImplementedError
+
+class ImageDetectionModel(DetectionModel):
+    """Parent image detection model class
+
+    :param model: Detection model object
+    :type model: Any
+    :param model_type: Model type (e.g. scripted, compiled, etc.)
+    :type model_type: str
+    :param model_cfg: JSON file containing model configuration (e.g. image size or normalization parameters)
+    :type model_cfg: str
+    :param ontology_fname: JSON file containing model output ontology
+    :type ontology_fname: str
+    :param model_fname: Model file or directory, defaults to None
+    :type model_fname: Optional[str], optional
+    """
+
+    def __init__(
+        self,
+        model: Any,
+        model_type: str,
+        model_cfg: str,
+        ontology_fname: str,
+        model_fname: Optional[str] = None,
+    ):
+        super().__init__(model, model_type, model_cfg, ontology_fname, model_fname)
+
+    @abstractmethod
+    def inference(self, image: Image.Image) -> List[dict]:
+        """Perform inference for a single image
+
+        :param image: PIL image
+        :type image: Image.Image
+        :return: List of detection results
+        :rtype: List[dict]
+        """
+        raise NotImplementedError
+
+    @abstractmethod
+    def eval(
+        self,
+        dataset: dm_dataset.ImageDetectionDataset,
+        split: Union[str, List[str]] = "test",
+        ontology_translation: Optional[str] = None,
+        predictions_outdir: Optional[str] = None,
+        results_per_sample: bool = False,
+    ) -> pd.DataFrame:
+        """Evaluate the image detection model
+
+        :param dataset: Image detection dataset for which the evaluation will be performed
+        :type dataset: ImageDetectionDataset
+        :param split: Split(s) to use, defaults to "test"
+        :type split: Union[str, List[str]]
+        :param ontology_translation: JSON file containing translation between dataset and model output ontologies
+        :type ontology_translation: Optional[str]
+        :param predictions_outdir: Directory to save predictions per sample, defaults to None. If None, predictions are not saved.
+        :type predictions_outdir: Optional[str]
+        :param results_per_sample: Whether to store results per sample or not, defaults to False. If True, predictions_outdir must be provided.
+        :type results_per_sample: bool
+        :return: DataFrame containing evaluation metrics
+        :rtype: pd.DataFrame
+        """
+        raise NotImplementedError
+
+class LiDARDetectionModel(DetectionModel):
+    """Parent LiDAR detection model class
+
+    :param model: Detection model object
+    :type model: Any
+    :param model_type: Model type (e.g. scripted, compiled, etc.)
+    :type model_type: str
+    :param model_cfg: JSON file with model configuration
+    :type model_cfg: str
+    :param ontology_fname: JSON file containing model output ontology
+    :type ontology_fname: str
+    :param model_fname: Model file or directory, defaults to None
+    :type model_fname: Optional[str], optional
+    """
+
+    def __init__(
+        self,
+        model: Any,
+        model_type: str,
+        model_cfg: str,
+        ontology_fname: str,
+        model_fname: Optional[str] = None,
+    ):
+        super().__init__(model, model_type, model_cfg, ontology_fname, model_fname)
+
+    @abstractmethod
+    def inference(self, points: np.ndarray) -> List[dict]:
+        """Perform inference for a single LiDAR point cloud
+
+        :param points: N x 3 or N x 4 point cloud array
+        :type points: np.ndarray
+        :return: List of detection results
+        :rtype: List[dict]
+        """
+        raise NotImplementedError
+
+    @abstractmethod
+    def eval(
+        self,
+        dataset: dm_dataset.LiDARDetectionDataset,
+        split: Union[str, List[str]] = "test",
+        ontology_translation: Optional[str] = None,
+        predictions_outdir: Optional[str] = None,
+        results_per_sample: bool = False,
+    ) -> pd.DataFrame:
+        """Perform evaluation for a LiDAR detection dataset
+
+        :param dataset: LiDAR detection dataset for which the evaluation will be performed
+        :type dataset: LiDARDetectionDataset
+        :param split: Split or splits to be used from the dataset, defaults to "test"
+        :type split: Union[str, List[str]]
+        :param ontology_translation: JSON file containing translation between dataset and model output ontologies
+        :type ontology_translation: Optional[str]
+        :param predictions_outdir: Directory to save predictions per sample, defaults to None. If None, predictions are not saved.
+        :type predictions_outdir: Optional[str]
+        :param results_per_sample: Whether to store results per sample or not, defaults to False. If True, predictions_outdir must be provided.
+        :type results_per_sample: bool, optional
+        :return: DataFrame containing evaluation metrics
+        :rtype: pd.DataFrame
+        """
+        raise NotImplementedError

docs/py_docs/_build/html/detectionmetrics.cli.html

@@ -74,30 +74,14 @@ <h1>detectionmetrics.cli package<a class="headerlink" href="#detectionmetrics-cl
 <section id="submodules">
 <h2>Submodules<a class="headerlink" href="#submodules" title="Link to this heading"></a></h2>
 </section>
-<section id="module-detectionmetrics.cli.batch">
-<span id="detectionmetrics-cli-batch-module"></span><h2>detectionmetrics.cli.batch module<a class="headerlink" href="#module-detectionmetrics.cli.batch" title="Link to this heading"></a></h2>
+<section id="detectionmetrics-cli-batch-module">
+<h2>detectionmetrics.cli.batch module<a class="headerlink" href="#detectionmetrics-cli-batch-module" title="Link to this heading"></a></h2>
 </section>
-<section id="module-detectionmetrics.cli.evaluate">
-<span id="detectionmetrics-cli-evaluate-module"></span><h2>detectionmetrics.cli.evaluate module<a class="headerlink" href="#module-detectionmetrics.cli.evaluate" title="Link to this heading"></a></h2>
-<dl class="py function">
-<dt class="sig sig-object py" id="detectionmetrics.cli.evaluate.parse_split">
-<span class="sig-prename descclassname"><span class="pre">detectionmetrics.cli.evaluate.</span></span><span class="sig-name descname"><span class="pre">parse_split</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">ctx</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">param</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">value</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#detectionmetrics.cli.evaluate.parse_split" title="Link to this definition"></a></dt>
-<dd><p>Parse split argument</p>
-</dd></dl>
-
+<section id="detectionmetrics-cli-evaluate-module">
+<h2>detectionmetrics.cli.evaluate module<a class="headerlink" href="#detectionmetrics-cli-evaluate-module" title="Link to this heading"></a></h2>
 </section>
-<section id="module-detectionmetrics.cli">
-<span id="module-contents"></span><h2>Module contents<a class="headerlink" href="#module-detectionmetrics.cli" title="Link to this heading"></a></h2>
-<dl class="py function">
-<dt class="sig sig-object py" id="detectionmetrics.cli.get_dataset">
-<span class="sig-prename descclassname"><span class="pre">detectionmetrics.cli.</span></span><span class="sig-name descname"><span class="pre">get_dataset</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">task</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">input_type</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">dataset_format</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">dataset_fname</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">dataset_dir</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">split_dir</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">train_dataset_dir</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">val_dataset_dir</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">test_dataset_dir</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">images_dir</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">labels_dir</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">data_suffix</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">label_suffix</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">ontology</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">split</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#detectionmetrics.cli.get_dataset" title="Link to this definition"></a></dt>
-<dd></dd></dl>
-
-<dl class="py function">
-<dt class="sig sig-object py" id="detectionmetrics.cli.get_model">
-<span class="sig-prename descclassname"><span class="pre">detectionmetrics.cli.</span></span><span class="sig-name descname"><span class="pre">get_model</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">task</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">input_type</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">model_format</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">model</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">ontology</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">model_cfg</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#detectionmetrics.cli.get_model" title="Link to this definition"></a></dt>
-<dd></dd></dl>
-
+<section id="module-contents">
+<h2>Module contents<a class="headerlink" href="#module-contents" title="Link to this heading"></a></h2>
 </section>
 </section>