Update NPU models and docs

docs/docs/configuration/hardware_acceleration_enrichments.md

@@ -5,7 +5,7 @@ title: Enrichments
 
 # Enrichments
 
-Some of Frigate's enrichments can use a discrete GPU / NPU for accelerated processing.
+Some of Frigate's enrichments can use a discrete GPU or integrated GPU for accelerated processing.
 
 ## Requirements
 
@@ -18,8 +18,10 @@ Object detection and enrichments (like Semantic Search, Face Recognition, and Li
 - **Intel**
 
   - OpenVINO will automatically be detected and used for enrichments in the default Frigate image.
+  - **Note:** Intel NPUs have limited model support for enrichments. GPU is recommended for enrichments when available.
 
 - **Nvidia**
+
   - Nvidia GPUs will automatically be detected and used for enrichments in the `-tensorrt` Frigate image.
   - Jetson devices will automatically be detected and used for enrichments in the `-tensorrt-jp6` Frigate image.
 

docs/docs/configuration/object_detectors.md

@@ -261,6 +261,8 @@ OpenVINO is supported on 6th Gen Intel platforms (Skylake) and newer. It will al
 
 :::tip
 
+**NPU + GPU Systems:** If you have both NPU and GPU available (Intel Core Ultra processors), use NPU for object detection and GPU for enrichments (semantic search, face recognition, etc.) for best performance and compatibility.
+
 When using many cameras one detector may not be enough to keep up. Multiple detectors can be defined assuming GPU resources are available. An example configuration would be:
 
 ```yaml

docs/docs/configuration/semantic_search.md

@@ -78,7 +78,7 @@ Switching between V1 and V2 requires reindexing your embeddings. The embeddings
 
 ### GPU Acceleration
 
-The CLIP models are downloaded in ONNX format, and the `large` model can be accelerated using GPU / NPU hardware, when available. This depends on the Docker build that is used. You can also target a specific device in a multi-GPU installation.
+The CLIP models are downloaded in ONNX format, and the `large` model can be accelerated using GPU hardware, when available. This depends on the Docker build that is used. You can also target a specific device in a multi-GPU installation.
 
 ```yaml
 semantic_search:
@@ -90,7 +90,7 @@ semantic_search:
 
 :::info
 
-If the correct build is used for your GPU / NPU and the `large` model is configured, then the GPU / NPU will be detected and used automatically.
+If the correct build is used for your GPU / NPU and the `large` model is configured, then the GPU will be detected and used automatically.
 Specify the `device` option to target a specific GPU in a multi-GPU system (see [onnxruntime's provider options](https://onnxruntime.ai/docs/execution-providers/)).
 If you do not specify a device, the first available GPU will be used.
 

frigate/detectors/detection_runners.py

@@ -161,12 +161,12 @@ class CudaGraphRunner(BaseModelRunner):
     """
 
     @staticmethod
-    def is_complex_model(model_type: str) -> bool:
+    def is_model_supported(model_type: str) -> bool:
         # Import here to avoid circular imports
         from frigate.detectors.detector_config import ModelTypeEnum
         from frigate.embeddings.types import EnrichmentModelTypeEnum
 
-        return model_type in [
+        return model_type not in [
             ModelTypeEnum.yolonas.value,
             EnrichmentModelTypeEnum.paddleocr.value,
             EnrichmentModelTypeEnum.jina_v1.value,
@@ -239,9 +239,30 @@ class OpenVINOModelRunner(BaseModelRunner):
             EnrichmentModelTypeEnum.jina_v2.value,
         ]
 
+    @staticmethod
+    def is_model_npu_supported(model_type: str) -> bool:
+        # Import here to avoid circular imports
+        from frigate.embeddings.types import EnrichmentModelTypeEnum
+
+        return model_type not in [
+            EnrichmentModelTypeEnum.paddleocr.value,
+            EnrichmentModelTypeEnum.jina_v1.value,
+            EnrichmentModelTypeEnum.jina_v2.value,
+            EnrichmentModelTypeEnum.arcface.value,
+        ]
+
     def __init__(self, model_path: str, device: str, model_type: str, **kwargs):
         self.model_path = model_path
         self.device = device
+
+        if device == "NPU" and not OpenVINOModelRunner.is_model_npu_supported(
+            model_type
+        ):
+            logger.warning(
+                f"OpenVINO model {model_type} is not supported on NPU, using GPU instead"
+            )
+            device = "GPU"
+
         self.complex_model = OpenVINOModelRunner.is_complex_model(model_type)
 
         if not os.path.isfile(model_path):
@@ -500,7 +521,7 @@ def get_optimized_runner(
             return OpenVINOModelRunner(model_path, device, model_type, **kwargs)
 
     if (
-        not CudaGraphRunner.is_complex_model(model_type)
+        not CudaGraphRunner.is_model_supported(model_type)
         and providers[0] == "CUDAExecutionProvider"
     ):
         options[0] = {