Fix label mapping issues

geti_sdk/data_models/containers/label_list.py

@@ -47,7 +47,7 @@ def _generate_indices(self):
         Map names and ID's to Label objects to enable quick label retrieval
         """
         self._id_mapping = {x.id: x for x in self.data}
-        self._name_mapping = {x.name for x in self.data}
+        self._name_mapping = {x.name: x for x in self.data}
 
     def get_by_id(self, id: str) -> Label:
         """

geti_sdk/data_models/label.py

@@ -114,7 +114,9 @@ class ScoredLabel:
 
     _identifier_fields: ClassVar[List[str]] = ["id"]
 
-    probability: float
+    probability: float = attr.field(converter=float)  # float converter here to make
+    # sure we're storing probability
+    # as a float64 dtype
     name: Optional[str] = None
     color: Optional[str] = None
     id: Optional[str] = None

geti_sdk/deployment/deployed_model.py

@@ -760,7 +760,7 @@ def get_model_config(self) -> Dict[str, Any]:
         config = {}
         for child in model_info_node:
             value = child.attrib["value"]
-            if " " in value:
+            if " " in value and "{" not in value:
                 value = value.split(" ")
                 value_list = []
                 for item in value:
@@ -769,6 +769,9 @@ def get_model_config(self) -> Dict[str, Any]:
                     except ValueError:
                         value_list.append(item)
                 config[child.tag] = value_list
+            elif "{" in value:
+                # Dictionaries are kept in string representation
+                config[child.tag] = value
             else:
                 try:
                     value = int(value)
@@ -802,6 +805,7 @@ def _clean_model_config(self, configuration: Dict[str, Any]) -> Dict[str, Any]:
             "model_type",
             "optimization_config",
             "task_type",
+            "labels",
         ]
         for key in unused_keys:
             configuration.pop(key, None)

geti_sdk/deployment/predictions_postprocessing/results_converter/results_to_prediction_converter.py

@@ -51,7 +51,7 @@ class InferenceResultsToPredictionConverter(metaclass=abc.ABCMeta):
     def __init__(
         self, labels: LabelList, configuration: Optional[Dict[str, Any]] = None
     ):
-        self.labels = labels
+        self.labels = labels.get_non_empty_labels()
         self.configuration = configuration
 
     @abc.abstractmethod
@@ -175,8 +175,12 @@ def __init__(
                 self.confidence_threshold = configuration["confidence_threshold"]
             if "label_ids" in configuration:
                 # Make sure the list of labels is sorted according to the order
-                # defined in the ModelAPI configuration
-                self.labels.sort_by_ids(configuration["label_ids"])
+                # defined in the ModelAPI configuration. If the 'label_ids' field
+                # only contains a single label, it will be typed as string. No need
+                # to sort in that case
+                ids = configuration["label_ids"]
+                if not isinstance(ids, str):
+                    self.labels.sort_by_ids(configuration["label_ids"])
 
     def _detection2array(self, detections: List[Detection]) -> np.ndarray:
         """